EMERGENCY MANAGEMENT IN SOCIAL MEDIA GENERATION
Deliverable 2.4 Impact of Social Media for ES and citizens using EmerGent concepts Version 1 (DoW, SP)
Thomas Spielhofer, Kerstin Junge, Joe Cullen, Anna Sophie Hahne, Cristina Castellanos and Matt Gieve (TIHR), Matthias Habdank and Chrstoph Amelunxen (UPB) and Marc-André Kaufhold and Christian Reuter (USI)
Tavistock Institute of Human Relations July 2017
Work Package 2
Project Coordinator Prof. Dr.-Ing. Rainer Koch (University of Paderborn)
7th Framework Programme for Research and Technological Development COOPERATION SEC-2013.6.1-1: The impact of social media in emergencies
D2.4: Impact of social media, Version V1, PU
Distribution level Public (P)
Due date 31/07/2017 (M40)
Sent to coordinator 28/07/2017 (M40)
No. of document D2.4
Title Impact of social media on Emergency Services and Citizens
Status & Version Final version
Work Package 2: Impact of Social Media in Emergencies
Related Deliverables D2.1, D2.2, D2.3, D2.5, D3.1, D3.6
Leading Partner TIHR
Leading Authors Thomas Spielhofer, TIHR (sections 1, 2, 3, 4.3, 4.4, 4.5, 5.1, 5.7, 7 and Appendix 3, 4 and 5) Kerstin Junge, TIHR (sections 5.2, 7, 8.2) Joe Cullen, TIHR (sections 4.1, 4.2, 8.1 and Appendix 1 and 2) Matt Gieve, TIHR (sections 5.5, 5.6 and 6.1) Anna Sophie Hahne (section 5.4, 6.1, and Appendix 9.) Cristina Castellanos (section 6.2) David Drabble (section 4.6 and 5.3)
Contributors Marc-André Kaufhold and Christian Reuter, USI (section 4.7 and Appendix 7) Matthias Habdank and Christoph Amelunxen. UPB (section 4.8 and Appendix 8)
Reviewers Joe Cullen, TIHR (internal) Christian Reuter, USI (external)
Keywords Social media, impact, emergencies, case study, tools, outcomes, Twitter, Facebook
This project has received funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement no 608352.
D2.4: Impact of social media, Version V1, PU Table of contents
List of Figures ...... IV List of Tables ...... V List of Abbreviations ...... VI Data Privacy, Protection and Ethical Checklist ...... VII 1 Introduction ...... 1 Abstract ...... 1 Purpose of the document ...... 1 Target audience ...... 2 2 Methodology ...... 3 Introduction ...... 3 Online surveys of citizens and emergency services ...... 3 Case study Wave 3: use of technologies ...... 5 Case studies of terrorist attacks: Boston and Brussels ...... 7 Case study Wave 3: evaluation of EmerGent ...... 8 Contribution analysis ...... 9 3 Results of the online surveys ...... 11 Online surveys of citizens ...... 11 Online survey of emergency services ...... 20 4 Organisational case study narratives ...... 25 Eindhoven Municipality, the Netherlands ...... 25 Social Rotterdam-Rijnmond Safety Authority, the Netherlands ...... 26 National Police, Unit East, Police Service Center, the Netherlands ...... 28 The Antwerp Fire Service, Belgium ...... 29 The National Crisis Center of the Federal Public Service, Belgium ...... 30 The Greater Manchester Police Service, UK ...... 31 Frankfurt Fire Service, Germany ...... 31 Graz Emergency Service, Austria ...... 33 5 Key findings of organisational case studies ...... 36 Social media tools, platforms or technologies ...... 36 Organisational structures and facilitators ...... 40
I D2.4: Impact of social media, Version V1, PU Social media users and uses ...... 45 Information validation ...... 48 Emergency service staff skills and resources ...... 49 Moderating citizen communities ...... 51 Lessons learned and implications for EmerGent ...... 53 6 Use of social media in response to terrorism ...... 55 The Brussels bombings ...... 55 The Boston bombings ...... 70 7 Case studies of the EmerGent trials ...... 79 Introduction ...... 79 Organisational backgrounds and use of social media ...... 79 Motivations for trialling the EmerGent system ...... 79 Trial design ...... 80 Results of the trials ...... 82 Learning ...... 82 Future use ...... 83 Concluding summary ...... 84 8 Conclusions ...... 85 Introduction: What is contribution analysis and what role does it play in EmerGent? ...... 85 Implementation of the Contribution Analysis ...... 86 Conclusions and Implications ...... 122 9 Summary ...... 125 Methodology ...... 125 Online surveys ...... 125 Organisational case study narratives ...... 126 Key findings and lessons learned ...... 126 Use of social media in response to terrorism ...... 128 Case study of the EmerGent system ...... 129 Conclusions ...... 129 References ...... 130 Appendix 1: Eindhoven Municipality (Case Study Report) ...... 134 Appendix 2: Rotterdam-Rijnmond Safety Authority (Case Study Report) ...... 140
II D2.4: Impact of social media, Version V1, PU Appendix 3: Netherlands, National Police, Unit East, Police Service Center (Case Study Report) ...... 150 Appendix 4: Antwerp Fire Service (Case Study Report) ...... 154 Appendix 5: The National Crisis Center of the Federal Public Service, Belgium (Case Study Report) ...... 163 Appendix 6: Greater Manchester Police (Case Study Report) ...... 168 Appendix 7: Frankfurt Fire Department (Case Study Report) ...... 178 Appendix 8: Graz Emergency Services (Case Study Report) ...... 191 Appendix 9: Significance Tables...... 199 Appendix 10: Full tables - Revised explanation tables ...... 202
III D2.4: Impact of social media, Version V1, PU List of Figures Figure 1: Use of social media during an emergency ...... 11 Figure 2: Types of information shared – Weather conditions/warning and feelings/emotions 12 Figure 3: Types of information shared – Videos and photos ...... 13 Figure 4: Types of information shared – other types ...... 14 Figure 5: Reasons for not using social media during an emergency ...... 16 Figure 6: Expected responsiveness of emergency services to messages posted via social media ...... 17 Figure 7: Likelihood of using an app in future for different purposes ...... 19 Figure 8: How frequently does your organisation currently use social media? ...... 22 Figure 9: Do you expect your organisation to increase its use of social media in future? ...... 23 Figure 10: Views on usefulness of different types of information shared on social media ...... 24 Figure 11: Eindhoven Location ...... 25 Figure 12 : The Rotterdam-Rijnmond Safety Region ...... 26 Figure 13: Structure of the Rotterdam-Rijnmond Safety Authority ...... 28 Figure 14: Map of regions in the Netherlands – East Region is number 2 ...... 29 Figure 15: Map of Antwerp port area ...... 30 Figure 16: Infrastructure of Feuerwehr Frankfurt (W3)...... 32 Figure 17: The structure of the ES in Styria in Austria ...... 34 Figure 18: the route of the amok driver...... 35 Figure 19: A trajectory of social media use ...... 45 Figure 20: Location of Brussels within Belgium ...... 55 Figure 21: Sites of the Brussels bombing of 2016...... 56 Figure 22: Location of the two explosions at Brussels Airport ...... 57 Figure 23: Global Density of Twitter Traffic – Brussels bombings ...... 58 Figure 24: Tweets from the information phase ...... 60 Figure 25: Typical tweet of phase of emotion ...... 60 Figure 26: Percentage of posts by language ...... 62 Figure 27: Number of tweets per hour...... 62 Figure 28: Prediction of sentiment per day ...... 64 Figure 29: Prediction of Sentiment per hour on 22nd March 2016 ...... 65 Figure 30: Prediction of Maxent per hour on 23nd March 2017 ...... 65 Figure 31: Word cloud of tweets on the 22nd between 10am and 11am ...... 66 Figure 32: Word cloud of tweets on the 22nd between 6pm and 7pm ...... 67 Figure 33: Word cloud of tweets on the 23rd between 12pm and 1pm ...... 67 Figure 34: Examples of “PorteOuverte” ...... 69 Figure 35: Location of Boston in the United States ...... 70 Figure 36: Boston and municipalities around ...... 71 Figure 37: Significant sites related to the Boston Marathon bombing of 2013...... 72 Figure 38: Cumulative time series of tweets from within a 35 mile radius of the Boston Marathon finish line selected using the stems “explode*”, “explos*” and bomb*” after the bombings at 2:49 p.m...... 74 Figure 39: Emergent Summary Theory of Change ...... 91
IV D2.4: Impact of social media, Version V1, PU List of Tables Table 1: ES staff use and attitudes towards social media ...... 21 Table 2: Use of social media in 2014 and now ...... 22 Table 3: Enablers and barriers to social media use in ES organisations ...... 43 Table 4: Number of messages per platform ...... 61 Table 5: Prediction Accuracy per machine learning algorithm ...... 64 Table 6: Comparison of the Emergent Trials in Dortmund and Hamburg ...... 81 Table 7: EmerGent Summary Theory of Change ...... 89 Table 8: Supporting and Refuting Evidence ...... 92 Table 9: Assumptions ...... 94 Table 10: Alternative primary explanatory mechanisms M1A ...... 96 Table 11: Alternative primary explanatory mechanisms M2A ...... 97 Table 12: Alternative primary explanatory mechanisms M3A ...... 98 Table 13: Influencing Factor - Cultural Contextualisation ...... 99 Table 14: Influencing Factor - Organisational Culture ...... 101 Table 15: Evidence collected – Mechanism 1 ...... 107 Table 16: Evidence collected – Mechanism 2 ...... 109 Table 17: Evidence collected – Mechanism 3 ...... 111 Table 18: Alternative Primary Explanation Mechanisms M1A (2) ...... 113 Table 19: Alternative Primary Explanation Mechanisms M2A (2) ...... 114 Table 20: Alternative Primary Explanation Mechanisms M3A (2) ...... 115 Table 21: Influencing Factor – Cultural Contextualisation (2) ...... 116 Table 22: Influencing Factor – Organisational Culture (2) ...... 117 Table 23: Key findings of the Contribution Analysis ...... 122
V D2.4: Impact of social media, Version V1, PU List of Abbreviations
Abbreviation Expression
A2C Authorities to Citizen
C2A Citizens to Authorities
C2C Citizen to Citizen
DOW Description of Work
EmerGent Emergency Management in Social Media Generation
EMC Emergency Management Cycle
ES Emergency Services
ESI Emergency Services Interface
FP7 EC 7th Framework Programme
ICT Information and Communication Technology
IG Information Gathering
IM Information Mining
IQ Information Quality
IR Information Routing
IT Information Technology
SM Social Media
ToC Theory of Change
VI D2.4: Impact of social media, Version V1, PU Data Privacy, Protection and Ethical Checklist
Data privacy, protection and ethics actions Comments
Does this deliverable, and the activities that Yes have fed into it, comply with the European Directives 95/46/EC and 2002/58/EC?
Does this deliverable, and the activities that Yes have fed into it, comply with relevant national data protection and privacy laws, codes of practice and Guidelines?
Has a data privacy and protection risk analysis Yes been carried out within the context of this deliverable (D.7.5, Table 2)?
If YES, were data privacy and protection risks Yes Use of Twitter messages identified? (specify risks identified) Quotes from individuals participating in interviews Case studies of organizations’ use of social media
If YES, have actions been identified to address Yes Permission sought from individuals these privacy and protection risks? (specify to publish Tweets actions identified) Anonymization of data Case studies of organizations agreed with research participants prior to publication
IF YES, have these data privacy and protection Yes actions been successfully implemented? (specify actions carried out)
Have details of the data privacy and Yes protection issues and actions taken been recorded in the Privacy Risk Register?
Has an ethical risks analysis been carried out Yes within the context of this deliverable (D.7.9, Table 7)?
If YES, were ethical risks identified? (specify No risks identified)
VII D2.4: Impact of social media, Version V1, PU
If YES, have actions been identified to address NA these ethical risks? (specify actions identified)
IF YES, have these ethical actions been NA successfully implemented? (specify actions carried out)
Have details of the ethical issues and actions NA taken been recorded in the Ethical Risk Register?
Are there any outstanding data privacy, data No protection and ethical issues relevant to this Deliverable?
VIII D2.4: Impact of social media, Version V1, PU 1 Introduction
Abstract This Deliverable builds on the work produced in D.2.2 [JSGC+14] and D2.3, which presented the results of the first and second round of case studies carried out in work package 2. Each of these focused on a particular type of emergency – D2.2 specifically on the riots that took place in London in 2011 and D2.3 on six examples of floods across Europe between 2010 and 2015 – and the way social media was used and how it impacted on those involved in the emergency. This Deliverable continues this work – via three successive and inter-related steps. The first part of this Deliverable (Section 3) presents evidence from a recent round of surveys exploring citizens’ and emergency service staff’s attitudes towards and use of social media in general and in emergencies based on online surveys conducted between November 2016 and June 2017. This includes nationally representative online surveys of UK citizens, Italian citizens, German citizens and Dutch citizens, as well as an opportunity sample of emergency service staff across Europe. The second part of this Deliverable (Sections 4, 5, 6 and 7) further extends the ‘multiple case study’ methodology of the previous rounds – consisting of three groups of case studies. The first group of case studies focus on the use of particular technologies to collect and analyse social media data in organisations, including emergency services. These case studies focused on five key themes that were highlighted in rounds one and/or two: social media tools, platforms or technologies, organisational structures and facilitators, social media users and uses, information validation, emergency service staff skills and resources and moderating citizen communities. The second group of case studies explore more recent examples of man- made emergencies – as presented in D2.2 with reference to the London riots – by exploring the use and impact of social media during the Boston marathon bombings in 2013 and, more recently, the Brussels airport and metro bomb attacks in 2016. Finally, the last and final round of case studies focussed on the testing and piloting of the EmerGent tool in Salzburg (February 2017), Dortmund (April-May 2017), and Hamburg (May-July 2017). The third part of this Deliverable (Section 8) presents the main conclusions from all five rounds of case studies and the surveys, focusing in particular on their implications for the EmerGent ‘Theory of Change’ and how the different aspects of the project contributed to bringing about different outputs, outcomes and impacts. The concluding section, Section 9, uses contribution analysis to assess the plausibility of EmerGent’s over-arching Theory of Change, and to review the evidence on its potential impacts. All case studies are provided in the Appendix (Appendices 1 to 8).
Purpose of the document This Deliverable contributes to Objective O1 of EmerGent – Analyse the impact of social media for citizens and for emergency services in the whole Emergency Management Cycle (EMC) today and tomorrow. The specific objective of this Deliverable is to develop and update the Theory of Change and hypotheses presented in D2.1 [CSJD+14], D2.2 [JSGC+15] and D2.3 [SJCG+16]. This is done via surveys exploring citizens (Task 2.3) and emergency services’ (Task 2.4) use of social media, and two rounds of case studies focusing on the use social media by
1/211 D2.4: Impact of social media, Version V1, PU particular emergency services or authorities to research the impact of social media on how emergency services responded and reacted to emergencies (Task 2.2) and how citizens groups or communities are engaged in this response, and what impact this has on them (Task 2.3).
Target audience The direct target audience for this Deliverable is the EmerGent project partners. Additional target audiences are emergency services, researchers in general and those with an interest in these particular case studies (social media analysis tools) and in social media use in emergencies in general, as well as the European Commission and other FP7 projects.
2/211 D2.4: Impact of social media, Version V1, PU 2 Methodology
Introduction Deliverable 2.4 builds on the results of the second wave of case studies, which focused on examples of the use and impact of social media in emergencies that involved flooding. These examples represented different locations and timeframes, i.e.: Germany (2013); Georgia (2015); Slovenia (2014); Poland (2010); Western Norway (2014); UK (2013-14). The objectives of Phase 3 are: i) to further deepen the project’s understanding of the use of social media in emergency situations and of the impacts associated with the use of social media and ii) contribute to revising and enhancing the EmerGent ‘Theory of Change’, developed in Task 2.1. This chapter presents the methodology of the various elements of this Deliverable as detailed in Section 1.1 above – including: The online surveys of citizens and emergency services (Section 2.2) The case studies of the use of social media technologies (Section 2.3) The case studies of terrorist attacks in Boston and Brussels (Section 2.4) The case studies of the EmerGent tool exercises and trials (Section 2.5) The contribution analysis of the overall EmerGent Theory of Change (Section 2.6).
Online surveys of citizens and emergency services Deliverable D2.3 [SJCG+16] presented the results of a survey of 1034 citizens (including 195 working or volunteering for an emergency service – excluded from the main analysis) across Europe conducted between February and June 2015. The main aims of the survey were to explore citizens’ attitudes towards the use of social media for private use and in emergency situations. However, a weakness of this survey was that it was based on an opportunity sample of citizens across Europe. Thus, project partners sent out a link to the online survey in English, Polish, Italian, German and Slovenian to friends, colleagues, professional and social contacts as well as via their own social media channels and websites. This meant that the sample of citizens responding to this survey was not fully representative, but was likely to be biased in favour of those with access to social media and also those living in partner countries. In order to collect more robust data and up-to-date information on citizens’ attitudes towards the use of social media for private use and in emergency situations, the research team decided therefore to commission four representative online surveys to be conducted in the UK, Germany, Italy, and the Netherlands. The surveys in Italy, the Netherlands and the UK were conducted by Opinium1 on behalf of the Tavistock Institute. The survey in Germany was carried out by GapFish2 on behalf of the University of Siegen. All surveys were conducted in the timeframe between October 2016 and March 2017. The surveys conducted in Italy, Netherlands and the UK were weighted to be in line with population statistics. The German survey used a sampling strategy to achieve representative responses from citizens with regard to their gender and age and therefore did not need to be weighted for these variables. Attempts at weighting it with regard to income of survey
1 http://opinium.co.uk/ 2 https://www.gapfish.com/en/
3/211 D2.4: Impact of social media, Version V1, PU participants were not possible due to lack of access to nationally representative data for this variable. The Dutch survey was weighted to be representative of the population regarding gender, age, household income, working status and region. National statistics were used from the Dutch Central Bureau of Statistic [WWW01]. The UK survey was weighted to be in line with population statistics regarding gender, age, region, working status and social grade. Population statistics were derived from the Office for National Statistics [WWW02]. In contrast to Germany, the Netherlands and the UK, Italy has a much lower proportion of internet users – 66%, compared with around 90% in the other three countries [WWW03]. For this reason in addition to gender, age, region, working status and household income, the data of the Italian survey was also weighted by Facebook and smart phone use as there was the risk that the data was biased towards internet users. National information about demographic statistics was used from the Italian National Institute of Statistics [WWW04] and information of Facebook and smart phone use was taken from a recent survey on internet and social media use across the world [Kemp17]. Deliverable 3.6, on the other hand, presented the result of a survey of 696 Emergency Service staff across 27 European countries conducted between 9th September and 15th October 2014. The main aims of the survey were to explore Emergency Service staff’s attitudes towards social media for private and organisational use and the levels and main factors influencing their current and likely future use of social media in their organisations. A follow-up survey of 473 emergency staff members (these included some who completed the first survey and shared their email addresses and others responded for the first time) between 24th April and 30th June 2017 in order to assess the extent to which attitudes towards social media have changed over the last three years since the beginning of the study. The analysis of all surveys was carried out using IBM SPSS Statistics 23. Tables and figures were created in Microsoft Excel. The analysis involved calculation of frequencies for all questions. To test for differences between the countries various significance tests were used depending on the level of measurement. Chi-Square tests3 were used for categorical variables, Kruskal-Wallis tests4 for ordinal variables and one-way ANOVAS5 for interval variables. Post hoc analyses6 were performed using the Scheffé post hoc criterion for significance. All significance tests were carried out with a significance level of .05. Statistical results including effect sizes, where possible, are presented in Appendix 9, these are indicated by a superscript letter. Effect sizes indicate the magnitude of an effect and are standardised so that they can be compared across different measures. Conventions for the interpretation of effect sizes exist so that they can be classified into small, medium and large effects. In addition to proportions per countries, figures are used to display combined results. The combined results take into account that for the citizen surveys the sample size is much lower in Germany (1069)
3 Chi-square test of independence are used to test if there is a significant association between two (mostly nominal) variables. 4 Kruskal-Wallis test is a non-parametric test for independent samples. It is used when the dependent variable is ordinal and there are more than two groups. 5 One-way analysis of variance (ANOVA) is used to test for significant differences between means of more than two groups. 6 As ANOVAS test omnibus hypotheses (i.e., there is a significant differences between at least two groups), post hoc tests are needed to determine which groups significantly differ.
4/211 D2.4: Impact of social media, Version V1, PU than in Italy, Netherlands and the UK (2001, 2001, and 2000 respectively) so that each country determined the combined results in equal measure.
Case study Wave 3: use of technologies The overall aim of the third wave of case studies focuses on researching the impact of social media on how ES respond and react in an emergency (Task 2.2) and how citizens groups or communities are engaged in this response, and what impact this has on them (Task 2.3). However, the primary focus of this round was on ES or authorities use existing tools or products (such as Brandwatch or Coosto) and how they interact with citizens before, during or after an emergency through, for example, moderating community-based social media volunteer groups. This third wave of case studies builds on the results of the previous wave to further explore a number of topics of interest for EmerGent that were identified through analysis of the ‘flooding’ cases. As with the two previous rounds of case studies, the methodology used in this third wave is based on a ‘multiple case study’ approach [Yin02] which allows for exploration of the impact of social media in emergency situations through the use of a ‘replication strategy’, in which successive case examples are selected to explore and confirm or disprove the patterns identified in the initial case examples. According to this model, if all or most of the cases provide similar results, there can be substantial support for the development of a preliminary theory that describes the phenomena. Essentially, this means ‘drilling down’ to explore in more detail themes and hypotheses that were highlighted as significant in the second phase of case studies. Six of these themes and hypotheses were identified, which the research questions the third wave of case studies addressed. The six themes were: Theme 1: Social media tools, platforms or technologies What types of social media tools, platforms or technologies are being used by ES? What are seen as the strengths and weaknesses of these? How are they integrated into the organisational structure of ES? Theme 2: Organisational structures and facilitators How is the use and analysis of social media integrated into the ES’s organisational structures and systems? To what extent does this facilitate or hinder the use of information gained from social media by these organisations? What ‘resistance to change’ dynamics in relation to the use of social media can be identified and how might these be addressed? Which organisational structures or arrangements are likely to facilitate the use of social media in emergencies?
Theme 3: Social media users and uses
5/211 D2.4: Impact of social media, Version V1, PU Which types of staff or other stakeholders are most likely to use social media during an emergency? What types of information are they most interested in: for example, situational awareness data from citizens or data on the public mood or the emergence of rumours or misinformation? What kinds of social media citizen roles would be most useful for ES? Theme 4: Information validation How can ES be persuaded that information provided via social media from citizens is credible and trustworthy? What procedures and tools are used to validate such information efficiently and effectively? How can reliability and accuracy be supported, technically? Theme 5: Emergency service staff skills and resources What skills are required to help ES staff to find, interpret and make use of social media data? What guidance or training would be most useful for them? Theme 6: Moderating citizen communities What approaches do ES already use to moderate and support volunteer communities using social media? How can moderating social media volunteer communities in emergencies be supported more efficiently and effectively? What can ES do before a disaster to make stronger links with such communities and what can they do to support the preparation, response and recovery operations? In Phase 3 of the case studies, the focus was on learning from real examples of the use of social media by emergency services. We were particularly interested in i) examples of emergency services that had embedded the use of social media in their organisational structures, systems and practice ii) specific platforms and technologies that were being used by emergency services iii) examples of real emergencies that illustrated i) and ii). The ‘unit of analysis’ for the Phase 3 case studies was therefore: a particular service; a particular platform; or a specific emergency. In other words the unit of analysis can be defined as ‘a socio- technical system’ that reflects the embedding and use of social media platforms and tools within an organisational environment and which was applied in emergency prevention, management and recovery. Learning from real examples marks a shift from the two previous case study waves – the first of which had an ‘exploratory’ and the second an ‘analytical’ emphasis – to applying transferable learning from good practices to the key EmerGent outputs. In this third round of case studies, the use of ‘benchmarking’ methods therefore assumes more significance than in previous rounds. In addition, we had a particular interest in this third wave of case studies in applying the concept of ‘value embedded action systems’ – outlined in D2.3 – to understand firstly how social media reflect the organisational culture and values of the emergency
6/211 D2.4: Impact of social media, Version V1, PU services who use them and, secondly, how these technologies become adapted and changed through experiences of their ‘use’. Previous case studies reported in D2.2 [JSGC+15] and D2.3 [SJCG+16] have shown that the use of social media is so far often only limited among ES for a number of practical and organisational reasons. For this round of case studies we wanted to therefore focus on ES or authorities known to use particular tools or approaches to analyse social media and to explore the organisational factors associated with the integration of such tools and how it impacted on their response. To this end, we explored links with existing contacts via the EAB and other networks to identify relevant examples. As a second step, we tried to contact emergency services identified on websites of particular tools or applications that could be used to collect and analyse social media (including Coosto, PublicSonar, Brandwatch, Musterpoint, and others). While this resulted in a few case study visits, in most cases we were not able to achieve agreement for participation – sometimes because we got no response to emails sent or calls made to these organisations, or because ES refused to participate in the research (this was particularly the case among police services who expressed concerns about discussing confidential information relating to their processes and procedures with the research team). This Deliverable therefore presents the results of the third case study wave that applied the methodology. Deliverable D2.4 also aims to explore the impact of the use of the EmerGent data-mining tool on organisations, to complement technical evaluation activities carried out as part of Work Packages 3 and 6.
Case studies of terrorist attacks: Boston and Brussels Previous case studies, with the exception of Deliverable D2.2, have largely focussed on natural emergencies and/or focussed on the emergency response in Europe alone. To fill this gap, the research team conducted two interrelated reviews of evidence to explore the impact of social media on how ES respond and react in an emergency (Task 2.2) and how citizens groups or communities are engaged in this response, and what impact this has on them (Task 2.3) with regard to two high-profile terrorist attacks: the Boston marathon bombing in the USA (2013) and the Brussels airport and metro bombings in Belgium (2016).
The Boston bombing The methodology of this case study consisted of three stages. The first phase consisted of a scoping stage to select an international case study which could complement the findings from previous case studies considering three main criteria: a focus on “man-made” emergencies, in contrast to “natural” emergencies (see Deliverable D2.3) existence of available literature on the event analysing or documenting the role of social media recentness of the emergency (no longer than 5 years ago, but ideally less), as the use of social media is changing constantly. After consideration of several possible cases, the Boston bombing was selected as it was seen as providing a good example of how social media can interact with local emergency management practices and plans previously prepared in a man-made emergency.
7/211 D2.4: Impact of social media, Version V1, PU Additionally, as a terrorist attack, some learning could be found that complements the analysis of the Brussels attack (see Section 2.4.2 below). The second phase comprised a review of the available documentation from the internet and in peer-reviewed article data bases. This was used to outline both the context and the background of events associated with the bombing in Boston’s Marathon and the following manhunt, as well as the analysis of the use of social media related to these events. Some direct checks in different social media such as Twitter, Facebook and Reddit were also carried out. Finally, an analysis and synthesis of the secondary research material was carried out, following the same research questions and guidelines as for the case studies conducted as part of D2.3
The Brussels terrorist attacks The methodology of this case study followed the approach described in Section 2.4.1 above with a few exceptions. In particular, the focus on the Brussels terrorist attack was instigated as a result of social media data collected by the research team on the day of the bombings – providing a rich source of data for further analysis. This involved two project partners collecting social media data at the same time, both using the free and public Twitter API. Twitter only gives you access to a certain proportion of messages over a limited period, therefore only a subset of all published activities were gathered. One project partner used the free software environment for statistical computing and graphics R [WWW05] to gather half a million tweets between 11:04:44 CET on the 22nd and 19:58:37 CET on the 23rd March 2016. No data from other Social Media platforms was collected and only English tweets were harvested that included at least one of the following hashtags: brussels, openhouse, porteouverte, brusselslockdown, brusselsexplosion, brusselsattack, brusselsairport, Maelbeek, Zavantem, IkWilHelpen. The other project partner was also constrained by public APIs but the data gathering was performed using NoSQL database [WWW06] for extracting and storing data. Data from Twitter, Instagram, Google+, Facebook and YouTube were harvested. Social Media that were collected had to include either ‘Brussels’ or ‘Bruxelles’ irrespective of the language of Social Media data. In total, nearly 3 million posts on Social Media were harvested between 11.44.20 CET on the 22nd March and 21.05.20 on the 23rd March 2016. The first phase of the case study therefore involved the use of a variety of statistical analysis techniques (described in Section 6.2) to determine the scale of information shared and to explore the sentiments of collected tweets and the change of sentiments over time. The second phase comprised a review of the available documentation from the internet and in peer-reviewed article data bases. This was used to outline both the context and the background of events associated with the bombings and the response. Finally, an analysis and synthesis of the secondary research material was carried out, following the same research questions and guidelines as for the case studies conducted as part of Deliverable D2.3.
Case study Wave 3: evaluation of EmerGent The set of case studies focusing on the use of the EmerGent system in the two field trials of 2017 (Dortmund and Hamburg) concludes the successive case study waves we have implemented as part of the EmerGent impact assessment. The unit of analysis for this case study round was the EmerGent system itself, and we worked with results and learnings from
8/211 D2.4: Impact of social media, Version V1, PU previous case study rounds covering the six themes outlined in Section 2.3 above to address two of the high level questions of the EmerGent impact assessment: what results has the EmerGent project achieved during its lifetime, and why?; what can we conclude from these results about EmerGent’s likely medium and long term impacts? To this end, the case studies were implemented using two of our case study methodologies. The bulk of data was collected with the help of qualitative interviews with ES staff in the participating organisations, covering a range of seniorities and functions. Where available, we also looked at documents relating to the trials, be this email exchanges or other documents produced. This data was then analysed using thematic content analysis and written up as a joint report to be able to reflect the combined achievements and lessons learnt.
Contribution analysis The Deliverable concludes with EmerGent’s ‘Contribution Story’. This uses contribution analysis to assess the plausibility of EmerGent’s over-arching Theory of Change, and to review the evidence on its potential impacts. Contribution analysis essentially involves constructing a plausible ‘contribution story’ that explains the contribution of a project, intervention or programme to it’s expected and/or identified outcomes and impacts. It has been described as ‘a pragmatic approach to applying the principles of theory-based evaluation. It assesses causal chains from beginning to end, reports on whether the intended changes occurred or not, and identifies the main contributions to such changes, including the intervention under evaluation’ [DeTo12]. Following standard practice [Mayn12] [BeMa14] the methodology applied to develop this contribution analysis for EmerGent involves six steps: 1. Set out the attribution problem to be addressed – specifying the outcome or target that is hoped to improve or change, as well as the key evaluation questions to be addressed 2. Develop a theory of change about how the intervention is supposed to work, together with i) the assumptions underpinning the theory ii) the risks to realisation of the intended outcomes and impacts 3. Gather evidence to assess whether the Theory of Change works, and explore and discuss plausible alternative explanations - identifying the most likely alternative explanations and the evidence associated with them 4. Assemble the Contribution Story – explain how and why a result is caused by a particular sequence of events and actions, and why it is reasonable to assume that the actions of the intervention have contributed to the observed outcomes. Specify the weaknesses in the story 5. Gather new evidence on the implementation of the intervention 6. Revise and strengthen the contribution story – using the new evidence gathered and assessed. Ultimately, Contribution Analysis aims to infer ‘plausible association’ between a project and a set of relevant outcomes by means of systematic inquiry. To demonstrate this ‘plausible association’, EmerGent’s embedded Theory of Change needs to meet the following five criteria Mayne [Mayn11]:
9/211 D2.4: Impact of social media, Version V1, PU Plausibility: Is the theory of change plausible? Implementation according to plan: Has the program been implemented with high fidelity? Evidentiary confirmation of key elements: To what extent are the key elements of the theory of change confirmed by new or existing evidence? Identification and examination of other influencing factors: To what extent have other influencing factors been identified and accounted for? Disproof of alternative explanations: To what extent have the most relevant alternative explanations been disproved? To address Criteria 4 and 5, we used the Relevant Explanation Finder (REF) [LBND12] which provides a guiding framework for the identification and examination of influencing factors and alternative explanations.
10/211 D2.4: Impact of social media, Version V1, PU 3 Results of the online surveys
Online surveys of citizens This section reports the findings from representative online surveys of citizens that were conducted in four European countries: Germany, Italy, Netherlands and the UK. A total of 7071 citizens of these four countries completed the survey. The main aims of the survey were to explore citizens’ attitudes towards the use of social media in emergency situations and to what extent they have already used social media in emergencies.
Use of social media in emergencies Analysis of the survey showed that almost half (45%) of citizens across the four countries have used social media during an emergency. However, as can be seen in Figure 1: Use of social media during an emergency, there are significant differences in the level of use across the four countries included in the survey (see Appendix 9 for a technical report providing details of significance tests carried out for each of the variables discussed in this section). Italy represents the country with the highest proportion of citizens that have used social media during an emergency. More than half of the Italian population (54%) has used social media during emergencies – 41% used it to share information. The lowest usage is reported in the UK, where only 36% of citizens have used it in the past to share and/or to find out information about an emergency. One possible explanation for this discrepancy is that Italy may have experienced more emergencies affecting large sections of the population (floods, earthquakes and avalanches) preceding the timing of this survey, while the survey in the UK was conducted in the time before the recent spate of terrorist attacks which saw very high levels of social media use. Figure 1: Use of social media during an emergency
100% 90% 80% 46% 55% 57% 55% 70% 64% 60% 50% 14% 40% 6% 20% 14% 16% 30% 14% 7% 6% 20% 5% 35% 6% 23% 10% 20% 21% 16% 0% Italy Germany Netherlands UK Total
Find and share information Share information only Find information only Not used
Note. N=6723 (348 missing).
Across all countries, citizens are most likely to share weather conditions or warnings via social media in emergencies (see Figure 2). This is the case especially in Germany, where nearly two-
11/211 D2.4: Impact of social media, Version V1, PU thirds (63%) of those who have used social media in this way used it to share weather conditions and warnings. In contrast, only 33% of citizens in Netherlands have shared this type of information. The second most likely information to share is feelings or emotions. Nearly half of the citizens in Germany (46%) have shared feelings or emotions and 31% of the UK citizens have shared this type of information.
Figure 2: Types of information shared – Weather conditions/warning and feelings/emotions
50% 46% 70% 63% 42% 39% 40% 60% 40% 48% 47% 50% 43% 31% 30% 40% 33% 30% 20% 20% 10% 10%
0% 0%
UK UK
Italy Italy
Total Total
Germany Germany
Netherlands Netherlands Weather conditions or warnings*** Your feelings or emotions***
Note. N=1983, *** indicates p < .001, ** p < .01 and * p<.05.
Another noticeable finding is that the proportion of Italians who have shared photos or videos (40% and 38% respectively) is considerably higher than for citizens across the four countries overall (see Figure 3).
12/211 D2.4: Impact of social media, Version V1, PU
Figure 3: Types of information shared – Videos and photos
50% 50% 38% 40% 40% 40%
30% 27% 25% 30% 21% 19% 18% 20% 16% 20% 15% 11% 10% 10%
0% 0%
UK UK
Italy Italy
Total Total
Germany Germany
Netherlands Netherlands A video*** An eyewitness photo***
Note. N=1983, *** indicates p < .001, ** p < .01 and * p<.05.
This indicates that there are differences between the countries in relation to what type of information citizens are more likely to share. Indeed, significant differences between the four countries were found for all types of information besides ‘Reassurance that you are safe’ and ‘Advice about what actions others should take to stay safe’ (see Figure 4).
13/211 D2.4: Impact of social media, Version V1, PU Figure 4: Types of information shared – other types
Note. N=1983, *** indicates p < .001, ** p < .01 and * p<.05.
14/211 D2.4: Impact of social media, Version V1, PU As was discussed above, a reasonably large proportion of citizens have shared information during an emergency, however most citizens have not. They might have done this deliberately or they may not have had the possibility or need to use social media during an emergency. So what are the barriers for using social media during an emergency? Figure 5 displays the main reasons why citizens would rather not use social media during an emergency. Reasons include concerns about credibility of social media content (such as false rumours or that information on social media is not reliable), technical concerns (that social media might not work in an emergency), as well as data privacy concerns (concerns about one’s own ability to use social media) and the general opinion that it is better to make an emergency call than to use social media for that purpose. Across all countries the biggest concern of citizens is that there are many false rumours on social media (63%).This fear of false rumours is of course not unjustified as there are examples of false rumours such as after the May 2017 Manchester suicide bombing when photos of supposedly missing people were posted on social media [Spie17] or the 2016 Brussels bombing where photos of a different attack were shared. This is particularly a concern for German citizens – 74% of them indicated that false rumours might put them off using social media during an emergency. The second biggest concern (57% of citizens overall) related to technical issues of using social media in emergencies. Only Dutch citizens seem to be less worried with only 49% expressing this technical concern as a reason for not using social media in an emergency.
15/211 D2.4: Impact of social media, Version V1, PU Figure 5: Reasons for not using social media during an emergency
Note. N=7071, *** indicates p < .001, ** p < .01 and * p<.05.
16/211 D2.4: Impact of social media, Version V1, PU Considering that many citizens have shared information during emergencies, do they have specific expectations from emergency services about using such data or responding to any requests for help? Overall, as can be seen in Figure 6 the survey showed that there is considerable variation between citizens in the four countries. In particular, more than two- thirds of German citizens expect their emergency services to regularly monitor social media, while in the UK only 37% of citizens share this expectation. Similarly, while 60% of Italian citizens think that emergency services should reply to any request for help sent via social media within an hour, in the UK the proportion was only 30%. This, combined with findings reported above (see Figure 5), indicate that at the time of the survey the majority of citizens in the UK did not expect emergency services to access information shared with them or others during an emergency. This contrasts with the views of citizens in the other three countries. Figure 6: Expected responsiveness of emergency services to messages posted via social media
80% 80%
60% 60% 31% 29% 23% 30% 40% 20% 40% 23% 21% 23% 17% 12% 20% 20% 36% 37% 34% 33% 33% 25% 32% 32% 27% 31%
0% 0%
UK
UK
Italy
Italy
Total
Total
Germany
Germany
Netherlands Netherlands Emergency services should regularly Emergency services are too busy to monitor their social media*** monitor social media during an emergency***
80%
60% 27% 40% 17% 21% 19% 20% 12% 33% 30% 27% 27% 18% Strongly agree
0% Agree
UK
Italy
Total
Germany Netherlands I would expect to get a response from them within an hour*** Note. N=7071, *** indicates p < .001, ** p < .01 and * p < .05.
17/211 D2.4: Impact of social media, Version V1, PU Downloading and using emergency apps The survey showed that the use of apps of relevance during emergencies is still in its infancy in most European countries – with the exception of the Netherlands where 28% of citizens had downloaded such an app. In contrast, only 7% of citizens in the UK, and 16% in Germany and in Italy had done so. There are different types of apps that could be helpful during an emergency: Weather apps, warning apps, first aid apps and emergency call apps. Across all countries weather apps (52%) were the most popular apps that citizens have downloaded. This proportion was especially high in Germany (69%), while it was the lowest in the UK with 44%. In the Netherlands warning apps were most common with 53% having downloaded such an app, while the average across all four countries was 42%. Downloading of emergency calls apps was more common in Italy (33%) and in the UK (28%) than in the Netherlands (18%) and in Germany (16%). Only the proportion of citizens that had downloaded a First Aid app did not vary significantly between the countries.
Although the majority of citizens across the four countries have not yet downloaded an emergency app, a larger proportion of respondents indicated that they were likely to use such an app in future. In particular, citizens across all countries thought that they were most likely to use such an app in future in order to receive emergency warnings – as can be seen in Figure 7, around 60% of citizens in Germany, Italy and the Netherlands stated that they thought they would use an emergency app for this purpose in future, whereas in the UK the proportion was considerably lower (39%). Differences between the countries were also found for all other reasons. The least popular reason for using an app in the future is to contact an emergency service. Especially in the UK only 21% would use an app for this purpose while in Italy 50% said they would do so.
18/211 D2.4: Impact of social media, Version V1, PU Figure 7: Likelihood of using an app in future for different purposes
Note. N=7071, *** indicates p < .001, ** p < .01 and * p < .05.
19/211 D2.4: Impact of social media, Version V1, PU Online survey of emergency services The Tavistock Institute conducted a survey of ES staff between April and June 2017 and received 473 survey responses across 25 European countries. As described in Section 2.2 above, the main focus of the survey was to explore the attitudes of ES staff towards social media and their views on its relevance and potential for the work done by their organisations and to compare and this with responses to a similar survey conducted in 2014 (see Deliverable 3.6). This section provides a summary of the main results of the survey, including: ES staff’s attitudes towards social media for private and organisational use Their current and likely future use of social media in their organisations The main factors enabling the use of social media by their organisations.
Attitudes towards the use of social media for both private and organisational purposes Table 1 below presents the results of a question asking ES staff to express their agreement or disagreement on a scale of 1 to 5 to a series of statements relating to their attitude towards social media for both private and organisational purposes. It shows that almost nine-in-ten of survey participants agreed that social media is useful for sharing information with citizens. Most ES staff also agreed the social media is useful for gaining situational awareness (78%) and for keeping in touch with the public (73%). It also shows that most of the ES staff surveyed used social media themselves (71%) in their private life. Comparing the results of this survey with the one carried out in 2014 shows a significant change (p<0.001) in response for six of the eight items. In particular, there appears to have been a significant change in attitude among ES with regard to the extent to which they see social media as an important tool in general and for gaining situational awareness during emergencies more specifically.
20/211 D2.4: Impact of social media, Version V1, PU Table 1: ES staff use and attitudes towards social media Statement Strongly Agree % Total agree Total agree agree % 2017 % 2014 %
Social media could be a 49 39 88 83 useful tool for emergency services to share information with citizens.
Most of my friends use social 37 46 83 76 media to keep in touch.
Social media could be useful 36 43 78 68 for gaining situational awareness information during emergencies.
It is important for emergency 34 40 73 68 services to use social media to keep in touch with the public during emergencies.
In my private life, I use social 34 37 71 66 media very often.
Social media are an 28 43 70 59 important tool for emergency services like the one I work for.
Information provided on 11 38 48 43 social media during an emergency is often not reliable.
Emergency services are too 11 22 33 27 busy to use social media
Due to rounding the percentages in the last column may be higher or lower than the sum of the 2nd and 3rd columns.
Current and future use of social media by emergency services Figure 8 below shows that ES staff reported that their organisations were currently most likely to use social media to share information with the public about how to avoid accidents or emergencies – however, only 37% of respondents said that they did this often, although
21/211 D2.4: Impact of social media, Version V1, PU 87% indicated that they did this at least occasionally. Just over half of respondents said that social media was used to receive messages from the public during emergencies at least occasionally and only 12% said this happened often in their organisations.
Figure 8: How frequently does your organisation currently use social media?
To share information about how to avoid accidents
To share information about how to behave in emergencies
For two-way communication with Often the public Sometimes/Occasionally
To gain situational awareness during emergencies
To receive messages from the public during emergencies
0% 20% 40% 60% 80% 100%
However, once again the use of social media has increased significantly for all five of these categories since 2014. This change is summarised in Table 2 below. It shows that the use of social media has increased for all five types of uses, although the change is largest for sharing information with the public – both about how to behave in emergencies and how to avoid accidents. Table 2: Use of social media in 2014 and now Type of use – at least occasionally: 2014 % 2017 %
To receive messages from the public during emergencies? 41 53
To gain situational awareness during emergencies 45 62
For two-way communication with the public 58 73
To share information about how to behave in emergencies 59 77
To share information about how to avoid accidents 68 87
Overall, the data shows that even though most respondents used social media occasionally for these different purposes, still less than half utilized it often, particularly during an
22/211 D2.4: Impact of social media, Version V1, PU emergency. In contrast, Figure 9 below shows that the majority of ES staff thought they expected their organisations to increase their use of social media in future, particularly in relation to sharing information with the public before and during emergencies. Thus, 87% of respondents thought that it was at least likely that their organisations would increase their use of social media to share information with the public about how to avoid accidents and a similar proportion would use it share information about how to behave during an emergency (85%). A lower proportion, but still around two-thirds (64%), also thought that their organisations would be more likely to utilize social media to receive messages from the public during emergencies. However, at the same time, analysis of the data suggested that overall those ES staff already using social media often in their organisations were significantly more likely to indicate that they expected their organisations to increase their use of social media than other respondents.
Figure 9: Do you expect your organisation to increase its use of social media in future?
To share information about how to avoid accidents
To share information about how to behave in emergencies
For two-way communication with the Very likely public Likely
To gain situational awareness during emergencies
To receive messages from the public during emergencies
0% 20% 40% 60% 80% 100%
This suggests that even though there is a willingness even among some currently rare users of social media in emergency services to expand their use of social media, the growth of this practice and the take-up of particular tools (such as the EmerGent system or Coosto) is likely to be greatest among, so called, ‘early adopters’ or ‘converts’ – who already use such technology and have the most positive attitudes towards social media. The survey also asked respondents the extent to which different types of information shared on social media by the public would be useful to them during an emergency. Figure 10 below shows that ES staff overall thought that photographs and general situational updates would be most useful to them. However, there were no significant changes in the types of information thought to be useful between 2014 and 2017. Furthermore, all types of
23/211 D2.4: Impact of social media, Version V1, PU information were most likely to be seen as useful by those who already used social media to receive or share information with the public often (or at least sometimes).
Figure 10: Views on usefulness of different types of information shared on social media
Photographs
General situational updates
Specific information Very useful Useful
Videos
Information about the public mood
0% 10% 20% 30% 40% 50% 60% 70% 80% 90%
Key factors to enable use of social media Finally, ES staff were asked what they saw as the main factors to ensure that social media is widely used by emergency services like their own. The most important perceived factors were: The skills of staff in using social media (44% thought this was ‘important’ and 42% ‘very important’ – 86% in total). The organisational culture of their service (40% thought this was ‘important’ and 44% ‘very important’ – 84% in total) The factors which were deemed to be less important were ‘funding for staff time’ (23% thought this was ‘very important’), the availability of equipment (30%) or software (34%) to access the data. It is also worth noting that compared with 2014, ES staff are now significantly more likely to rate staff skills (p<0.05), organisational culture (p<0.01) and software (p<0.01) as important to ensure the use of social media by their organisations. This suggests a growing recognition that tools such as the EmerGent system are needed to collect and analyse social media data – and that doing this relies on a supportive organisational culture and skills.
24/211 D2.4: Impact of social media, Version V1, PU 4 Organisational case study narratives This section presents the case study ‘narratives’, focusing on the situation in each of the examples studied. As noted above in Section 2, the third wave of case studies concentrates on researching the impact of social media on how ES respond and react in an emergency, with a particular interest in how emergency services use existing tools or products (such as Brandwatch or Coosto) and how they interact with citizens before, during or after an emergency. Unlike the previous round of case studies, in which the different case examples were linked together by a common theme - ‘flooding’ - this third round of cases discusses how social media has been used by eight specific emergency services and agencies responsible for public safety, in seven different European cities. The eight cases are: Eindhoven Municipality, the Netherlands Rotterdam-Rijnmond Safety Authority, the Netherlands National Police, Unit East, Police Service Center, the Netherlands. Antwerp Fire Service, Belgium Brussels Crisis Centre, Belgium Greater Manchester Police, UK Graz Emergency Services, Austria Frankfurt Fire Department, Germany The background situation in each of these cases is presented below.
Eindhoven Municipality, the Netherlands Eindhoven is a city in southern Holland, a key centre of the Dutch electronics and high-tech sector. The metropolitan area consists of 419,045 inhabitants and the city region has a population of 753,426 (Figure 11).
Figure 11: Eindhoven Location
25/211 D2.4: Impact of social media, Version V1, PU Eindhoven generates €24 billion of GDP and €55 billion in exports, one-quarter of the Dutch total. It absorbs 36% of all private Dutch R&D spending and is home to globally recognized companies including Philips, the healthcare, lighting and consumer product giant, and ASML, maker of photolithography equipment for the production of silicon chips. 18% of all Dutch automotive jobs are in Eindhoven, and 9% of all life technology employment. The Eindhoven University of Technology, with more than 7,000 students, is one of the top three research universities in Europe and its High Tech Campus founded by Philips houses over 80 companies employing another 7,000 residents Given its prominence as a high-tech hub, it comes as no surprise that Eindhoven Municipality has been a leading exponent of the ‘smart city’ movement and was an early pioneer in finding ways of applying information – especially ‘big data’ – to solve a range of problems, from traffic grid-lock, through developing the ‘circular economy’, to smart policing. Recent thinking around the ‘smart city’ has seen a shift in focus - from developing technologies that impose ‘order from above’ on citizens to create the perfect urban environment - to engaging with citizens as ‘co-producers’ of the knowledge required to make cities become the setting for a smart society. This has led Eindhoven to explore how social media can make a contribution. In line with other Dutch cities – like Rotterdam – Eindhoven’s interest in social media reflects a broader objective of working with ways of delivering ‘open government’.
Social Rotterdam-Rijnmond Safety Authority, the Netherlands The Rotterdam-Rijnmond Safety Authority is a governmental organisation which carries out duties on behalf of the region’s municipalities in the areas of disaster relief, crisis management, risk management, fire services, ambulance and medical assistance (Figure 12). Rotterdam-Rijnmond is one of 25 safety regions in the Netherlands.
Figure 12 : The Rotterdam-Rijnmond Safety Region
26/211 D2.4: Impact of social media, Version V1, PU A safety region is defined as ‘a geographical area within which governing bodies and operational services co-operate in terms of fire prevention, calamity response, crisis control and medical assistance in order to maintain and increase the safety of the inhabitants’. Its objectives are:
To improve protection for inhabitants against safety risks To improve the help offered during, and after incidents, hazards and crises To improve fire-prevention, fire-repression, medical assistance at accidents and disasters, risk management and centralised regional crisis management To increase overall managerial and operational performance The Rotterdam-Rijnmond Safety Authority covers a large and complex environment, including urban areas with 1.2 million inhabitants, a diverse infrastructure with a global harbour and 32.000 sea-going vessels, 110,000 inland vessels and around 800.000 shipping movements. This complexity is highlighted by the fact that the Safety Authority has over 80 Contingency Plans in place for various emergencies. 7 Reflecting this complexity and diversity, another distinguishing feature of the Safety Authority is its large network structure and organisational depth. As Figure 13 shows, the Safety Authority is a multi-stakeholder collaborative structure, involving 15municipalities in the Rijnmond area; emergency services; other public agencies (Regional public broadcasting: – RTV Rijnmond, Public Transport – RET, Ministry of Defence, Water Boards, Royal Dutch Rescue Service, Health Services) and commercial organisations, including energy companies and 60 companies in the harbour and industrial areas in cooperation with the municipality of Rotterdam.8 Co-ordination and strategic governance is provided through a ‘Common Regulation’ that incorporates four administrative bodies: Governing Board Executive Board Safety Region President Safety directorate.
7 Source: ‘Working for Safety’. Presentation, Rotterdam-Rinjhold Safety Authority, 2013. 8 Source: op. cit.
27/211 SRR: a large networkD 2.4: Impact of social media, Version V1, PU
Figure 13: Structure of the Rotterdam-Rijnmond Safety Authority
The Authority operates from a central office overlooking the port of Rotterdam, where it also co-ordinates a General Emergency Control Room for Police, Fire brigade and Ambulance Services (112). In ‘crisis mode’, a ‘crisis management team’ from within the Safety Region will co-ordinate the response to an emergency. It should be borne in mind that around 1,500 emergencies are logged every week within the Safety Authority region and only 1 or 2 of these need to be handled by the crisis management team.
National Police, Unit East, Police Service Center, the Netherlands The Unit East Police Service Centre of the Dutch National Police operates within the largest Unit of the Netherlands, (8.299 km2) serving 3.143.000 citizens and employing 6800 police officers. The Service Centre of the National Police is accessible to citizens via a direct telephone number (09008844) as an alternative for the emergency number (112) to contact the police for non-emergency information or enquiries. The Service Centre employs 220 agents, and has a staff of 21 members (10 Operational Experts and 11 Seniors) and is located in four locations within the Unit.
28/211 D2.4: Impact of social media, Version V1, PU
Figure 14: Map of regions in the Netherlands – East Region is number 2
The Antwerp Fire Service, Belgium The Antwerp Fire Service (AFS) or Brandweer Zone Antwerpen covers the city, the port and two communities – with around 600k inhabitants. Antwerp is the second largest port in Europe and is regarded as a high risk area due to the high concentration of petro-chemical facilities; there is a lot of construction and it is generally expanding a lot as a city. There are eight fire stations in Antwerp, including three in the port area and five in the city area. These used to be quite old fashioned, but have been increasingly modernised over the last 15 years, so they have been updated a lot.
29/211 D2.4: Impact of social media, Version V1, PU
Figure 15: Map of Antwerp port area
The AFS employs 700 operational staff who are all full professionals (no volunteers) and 80 support staff (lawyers, communications, finance, HR, etc.). It is an independent service – i.e. it is financed by the city, but it is ‘independent’ of it; however, the mayor is on the board of the fire service, so he is ‘our real boss’ as 80% of funding comes from the city. It deals with around 22,000 interventions per annum – this includes 10% of fires and 15% of floods – the rest is made up all other kinds of technical interventions, including car accidents, supporting the ambulance service, industrial accidents in the port, etc. More recently they have also had to focus on dealing with the aftermath of terrorism. There has been no major attack in Antwerp, although there have been incidents (arrests) and it is regarded as a high risk area as one of the key extremist groups was founded in Antwerp and alongside that Antwerp has a very large Jewish community.
The National Crisis Center of the Federal Public Service, Belgium The Belgium National Crisis Center of the Federal Public Service (FPS) Internal Affairs provides its services to the Federal Government and local municipalities in dealing with and responding to crises. One of its main tasks is to collect, analyse and disseminate the necessary information to the political and executive authorities. The Crisis Center also offers its infrastructure and know-how in interdepartmental management and crisis coordination at the national level. During a crisis, the Crisis Center will provide advice on what, when and how to communicate information to the public and discuss the overall communication strategy with the crisis team. The communication service of the Crisis Center employs seven staff full-time (up from four staff in March 2016) and can also draw on the support of a wider team of experts (Team D5) based in local municipalities – who can provide support for crisis communication. The Crisis Center can also draw on the services of a contact call-centre which citizens can call to share and receive information during a crisis.
30/211 D2.4: Impact of social media, Version V1, PU On 22nd March 2016, the Crisis Center was hosting a planned crisis communication exercise with the support Team D5 – which meant that 25 members of the Team D5 were on the way to the Brussels head-office at 8am, just as the attacks at the airport started on that day (see Section 6.1 for a detailed description of the terrorist attack).
The Greater Manchester Police Service, UK Greater Manchester is a metropolitan county in North West England, with a population of 2.56 million (the third most populous conurbation in the UK). It encompasses one of the largest metropolitan areas in the United Kingdom and comprises ten metropolitan boroughs. Greater Manchester Police (GMP) is the second largest Police Force in England and Wales and has around 6,000 police officers and 3,000 police staff. This is a reduction of 2,000 staff from a few years ago. GMP is organised into 10 policing territories, covering the 10 local authority areas. Overlapping these area Forces, GMP also has nineteen Branches responsible for Force Strategies and the delivery of services to front line policing [WWW07]. Whilst social media had been used prior to 2010, before then GMP did not have a social media strategy, but the intention was to use social media to make GMP more accessible. Social media was adopted in the context of a national conversation about police related issues which GMP wanted to be a part of. The newly appointed Web Manager did 6 months of research before implementing the strategy.
Frankfurt Fire Service, Germany Since 1874, Frankfurt am Main (Germany) has had a professional fire department, the Feuerwehr Frankfurt am Main, one of the biggest and most modern professional fire departments in the world. Employing approximately 1,000 people, this fire department is also one of the biggest public offices in the city administration of Frankfurt am Main (W1). With a population of 732,000, Frankfurt am Main is both the biggest city in the German federal state of Hessen and the fifth-largest city in Germany (W2). Inhabiting around 2.2 million people, the independent town is at the centre of the urban agglomeration Frankfurt- Rhine-Main. Around 2.5 million people (2012) live in the extended city region Frankfurt while the whole metropolitan region Rhine-Main has a population of approximately 5.5 million people. Twelve fire and rescue stations, organized in four groups (Figure 16, red marks 1-4), are spread throughout the urban area.
31/211 D2.4: Impact of social media, Version V1, PU
Figure 16: Infrastructure of Feuerwehr Frankfurt (W3) The fire officers are working 24/7 to protect the population of Frankfurt (W1). Besides fire engines and rescue vehicles that are used in almost every operation, several special vehicles and some extra equipment are available at the different locations. Firefighters that obtained a subject-specific training use these special vehicles and equipment to provide efficient and specialized assistance in unusual operations. This includes, for example, high angle rescue, rescue dog units, a technical assistance service, an environmental protection platoon or divers of the fire department. Both the professional fire department and the volunteer fire department, the latter including approximately 900 active members in 28 volunteer fire departments (Figure 16, blue dots) in the different city districts, are the main pillars of fire protection, rescue service and technical danger prevention. In Frankfurt, the fire department coordinates every non-police hazard prevention unit of the German charities that support civil protection: Arbeiter-Samariter-Bund Deutschland (independent charitable aid agency) Deutsches Rotes Kreuz (German Red Cross) Malteser Frankfurt (Catholic voluntary humanitarian organisation) Johanniter-Unfall-Hilfe (Protestant voluntary humanitarian organisation) Deutsche Lebens-Rettungs-Gesellschaft (German Life Saving Association) Technisches Hilfswerk (Federal Agency for Technical Relief) Furthermore, a youth and a mini fire department and the Feuerwehr- und Rettungsdiensttrainingscenter (Training Center for Fire Departments and Rescue Services) are affiliated to the Feuerwehr Frankfurt. Consequently, the central control room of Frankfurt is the biggest control room in the federal state of Hessen (W1). From here all
32/211 D2.4: Impact of social media, Version V1, PU firefighting and rescue operations as well as all hospital transports of the urban area are coordinated. It is stationed in the Brandschutz-, Katastrophenschutz- und Rettungsdienstzentrum (Center for Fire Protection, Civil Protection and Rescue Service). There are twelve equivalent work places whereof seven are staffed during the day and five at night. The other five work places serve in exceptional situations and as a failure reserve. Four monitors (table of operations, operation documentation, geographic information system, office communication) and a touchscreen to control different functions are available at every workplace. The operation administrators process around 800 to 1000 requests for assistance per day. Unfortunately, more than 50 % of the 112 calls are not emergency calls. The main reasons for these false calls are people dialling a wrong number, children playing with a mobile phone or even jokes. However, the administrators must not only process the requests for assistance via the emergency hotline but also around 2,500 additional administrative phone calls per day. Further, they are responsible for forwarding different concerns to the relevant departments. Annually, there are around 100,000 alerts concerning rescue and ambulance services and approximately 8,000 assistance services and 6,000 fire operations are necessary.
Graz Emergency Service, Austria Graz is located in the state of Styria in Austria. With a total area of 127.58km² and a population of 315.464 in 2016 it is the second-largest city after Vienna [WWW08]. The city itself is divided into 17 districts. The first district is called “Innere Stadt” and is part of the Old Town, which was declared a UNESCO World Heritage Site in 1999. It was in this part of the city that an incident involving a car ‘running amok’ mostly took place. It started in the 5th district, called Gries. In Figure 18 the numbers 1 to 3 are in Gries while 4 to 9 are located in Innere Stadt. As the incident took place at 12.25 am on the 20th of June 2015, which was a Saturday, the streets of the city centre were pretty crowded. In addition to a normal Saturday there were some events related to the 2015 Austrian Grand Prix. In Styria, the Emergency Services are divided in three parts as shown in Figure 17: The fire brigade, the ambulance service and the civil protection agency. Each of them has their own emergency number and their own control centre. The fire brigade in Styria consists of 692 volunteer fire brigades, 83 volunteer plant fire departments and one fire service in Graz. The ambulance service is called ÖRK (Österreichisches Rotes Kreuz / Austrian Red Cross). In case of a big crisis there is another authority called special department for civil protection and national defence (Fachabteilung Katastrophenschutz und Landesverteidigung). If this authority needs to be activated it can also use the resources of the Austrian militia (Bundesheer).
33/211 D2.4: Impact of social media, Version V1, PU
Figure 17: The structure of the ES in Styria in Austria
In case of a crisis the command structure depends on the extent of the incident. The highest rank in a city is the mayor of the city. Therefore he will have the power of order. If more than one city is affected the district commissioner takes over. If more than one district is affected by the incident, the next step in the hierarchy gets involved. This is the state of Styria with the special department for civil protection and national defence, as mentioned above. It is important to know that on the 20th June 2015 an event for the high-level firefighters was being held. Therefore they were all gathered in the same place when they heard about the incident. As the mayor of Graz was involved in the incident himself, he had made an emergency call and called his colleague in the municipal administration. In the beginning the situation was quite unclear. The people at the event did not know exactly what had happened; they just knew something big happened. The keywords reported were “car”, “shooting”, “knife attack” – which does not seem to fit together. It was said that there were a lot of injured and dead people. So the firefighters and politicians who were from Graz left the event and drove home. On the 20th June 2015 at 12.25am a 26year old Austrian started an amok by car. The driver attacked pedestrians and cyclists with his SUV (Daewoo Rexton) with an estimated speed of up to 100km/h (62mph). He killed 3 people (one infant, a 25year old woman and a 28year old man). 34 people were severely injured. The driver did not only use his car to attack people: at one point he got out of the car to attack two pedestrians with a knife. Figure 18 shows the route the driver took as well as all the places where he injured or killed people.
34/211 D2.4: Impact of social media, Version V1, PU
7 8
9
6
1 2 4 5 3
Figure 18: the route of the amok driver
No. action place
1 Starting point
2 Car hits 3 people – one dies Zweiglgasse-Lagergasse
3 Car hits 3 more people Augartenbrücke
4 Driver stops, leaves the car to attack Grazbachgasse 2 people with a knife
5 Car hits 2 cyclists Schönaugasse
6 Car hits 1 cyclist Dietrichsteinplatz & Schlögelgasse
7 Car hits several people – two die Hauptplatz Car crashes info a cafes outside terrace* – 8 people injured
8 Car hits 2 more cyclists Albrechtgasse
9 Driver surrenders to the police Police station in Schmiedgasse
* at the café there was an event related to the 2015 Austrian Grand Prix held
35/211 D2.4: Impact of social media, Version V1, PU 5 Key findings of organisational case studies This chapter draws out the main findings from these eight case studies (see Appendices 1-8) conducted as part of Task 2.2 and 2.3 with regard to each of the six themes presented in Section 2.3 above, building on the analysis of the use of social media and its impact and consequences for ES and citizens.
Social media tools, platforms or technologies The case studies revealed that there is a large variety of tools available for collecting and analysing social media, although most of these tools are aimed at commercial organisations rather than at emergency services. Some are available free of charge, or as demo versions, but most are subscription-only services. These include: TweetDeck9 is a free multi-column app for viewing Twitter feeds owned by Twitter – it can be set up to display a user’s Twitter timeline, mentions, direct messages, lists, trends, favourites, search results, hashtags, or all tweets by or to a single user. It is used by Greater Manchester Police and the Frankfurt Fire Department. They use it to monitor Tweets sent to their official accounts and those using particular keywords and hashtags. Coosto10 is a subscription-only service developed in the Netherlands supporting a wide range of social media applications, including supporting customer services management, publishing, and collecting and analysing social media content. It is used to collect, filter and analyse content from social networking sites, blogs and forums. It provides Boolean search operators and cross-reference searching; filtering; geolocation analysis and data export. It is mainly used by commercial organisations, but also by a few emergency services and public authorities, including four of our case studies: Eindhoven Municipality, Rotterdam Safety Authority, Antwerp and the Belgium Crisis Centre. PublicSonar11 is a subscription-only service also developed in the Netherlands aimed at providing a social media management and analysis tool for the public sector, including emergency services – Antwerp Fire Service were considering using it instead of Coosto as it was seen as more relevant to emergency services, including the provision of databases of local place names, and real time social media messages Musterpoint12 is a subscription-only service developed in the UK that also functions as a social media monitoring, management and response service also aimed specifically at public services, including emergency services. It is currently used by Surrey Police and Avon Fire & Rescue Service; no case studies were conducted of this tool CrowdControlHQ13 is a subscription-only service developed in the UK promoted as a ‘social media risk management and compliance platform’, and like Coosto can be used as both a social media management and analysis tool – according to its website, it is currently used by
9 https://tweetdeck.Twitter.com/ 10 https://www.coosto.com/en/ 11 http://publicsonar.com/ 12 http://www.musterpoint.co.uk/ 13 https://crowdcontrolhq.com/
36/211 D2.4: Impact of social media, Version V1, PU Cheshire Fire & Rescue Service [WWW09] and Wiltshire Police [WWW10]; and also by Greater Manchester Police Hootsuite14 is a subscription-only social media management and analysis application with a particular focus on the commercial sector – it was used by the Belgium Crisis Centre alongside TweetDeck and Coosto to monitor social media messages, and previously by Eindhoven Municipality before moving over to Coosto Signal15 is a cloud-based platform developed in New Zealand providing an intelligence monitoring, analytics and notification solution for open source social media content SocialSignIn16 is a similar European subscription-only service developed in the UK which according to its website is currently used by Staffordshire Police While most of these services are advertised as collecting and analysing social media data from a variety of sources, including Twitter, Facebook, blogs, and other sources, in practice, most of the case study sites visited focussed on information provided by Twitter and to some extent Facebook. The main reason for this is that these are used extensively during emergencies and are open (Twitter) or semi-open (Facebook) sources and therefore accessible for such tools. As noted though by several respondents, there is an awareness that other types of social media are growing in popularity, particularly among young people – and that such tools are often not able to capture information from such sources. This is not a major issue at the moment as there are still very large quantities of messages sent via Twitter and Facebook – up to 2 million for example during the Brussels bombing. Some of the case study respondents did use other types of social media – but this was mainly only to communicate with the public, rather than to collect information from them. Greater Manchester Police, for example, have over 100 social media accounts on platforms like Facebook, Instagram, YouTube, Flickr, AudioBoo, Kick, LinkedIn, Pinterest, Google+, Snapchat, and GoBubble. Each of these platforms are used for specific functions, such as GoBubble being used for kids during preventative work. Other services mainly used Twitter to send out information to the public before, during or after an emergency. [More details]
How are such tools used by emergency services and authorities? Four of the eight case studies (Frankfurt, Graz, Netherlands East Region, and GMP) used free tools only to access social media data. Greater Manchester Police (GMP) use TweetDeck predominantly to monitor particular social media accounts before, during or after an event or emergency. This is used to identify particular issues or events which may not have been picked up by normal channels. This included, for example, a tweet sent to the GMP Twitter account (@GMPolice) by the victim of a hate crime, complaining that it had not been dealt with by local police forces for several days. Further checks by the digital service team showed that no officer had been allocated to the crime and that the crime was labelled in the records as an assault, despite it being called a hate crime on Twitter. Similarly, the Netherlands East Region Police Service Centre use public platforms such as TweetDeck or Hootsuite In the case of a major incident or emergency to monitor what is going on: ’to find
14 https://hootsuite.com/en-gb/ 15 http://www.getsignal.info/ 16 https://socialsignin.net/
37/211 D2.4: Impact of social media, Version V1, PU out the different hashtags being used and to follow them’. And any information found in this way is shared with colleagues in the emergency rooms: ‘Have you seen this? – so that they can take any action if needed’. Four of the eight case studies used Coosto to collect and analyse social media data in a more sophisticated way. In Antwerp, the social media manager runs Coosto on her laptop continuously and will look at it ‘every few minutes’. This is done by setting up a very general search for any bits of social media information that could be relevant to an emergency, using very general search terms such as accident (crash, cars, trucks, dangerous), fire, explosion, flood, storms, and terrorism. This is then combined with Antwerp district or place names. To make sure they do not miss any posts, they also use slang and dialect words to denote particular areas. “We have built up new terms over the years – every time something picks up you add it. You need to think very carefully about search terms. Coosto then collates the info – we use it in dashboards – this links with the search in real time”. They also use Coosto to follow particular accounts on Twitter, such as “the Port of Antwerp” and use filters to exclude some false positives. This includes, for example, mentions of the term ‘Brand’ in messages – in Dutch this means ‘fire’ – in order to exclude any posts written in English talking about particular ‘brands’. Their searches still result in a lot of false positives, particularly in relation to the term ‘fire’: “With Coosto we don’t filter out ‘fire’ too much so we get them from all over - but as this is a serious event we’d rather get too many than too few”. If the Antwerp Fire Service social media manager spots several tweets about an event or emergency at the same time, she will check it and pass on any relevant information, including videos or photos, to the dispatch centre. “If we get extra information about an incident we would send the photos to the officer”. Eindhoven Municipality and Rotterdam Safety Authority use Coosto in a similar way. In Eindhoven, Coosto is used for a wide range of purposes – not just emergencies. For example, recently there was a major power supply failure. Coosto was used to monitor social media messages to identify citizens’ concerns about what was happening and when supply would be resumed. Another example is ‘Kings Day’ – the most important holiday in the season: “We’d just started using social media proactively, when we noticed a video of major abuse in the party area going viral on social media. We had to decide how to communicate this public affair. Our analysis based on Coosto information helped reach decisions for our communication strategy”. Most of the social media information fed into Coosto comes from Twitter. As is the case with the Rotterdam Safety Authority Safety Authority, Eindhoven Department of Communication works closely with traditional media – particularly the local newspaper, which is mainly published on-line, to channel and reinforce its messages to citizens and to help verify information. In contrast, the Crisis Centre in Belgium uses Coosto to analyse social media data during a major emergency or crisis but not on a day-to-day basis. During the Brussels bomb attacks on 22nd March 2016, for example, the National Crisis Centre used tweets by the public and other sources to gain useful situational awareness of the crisis as it unfolded. So, for example, through looking at what people were tweeting about, it was possible to pick up trends in what people were concerned about at that particular moment: ‘What are at this moment, the key concerns of citizens? What are the main questions right now that are bothering them? What is trending on social media?’ For example, at one particular moment when the transport network was shut down there were many messages sent on social media about ‘mobility’ – how to get home or where they were going. Or at about 12 o’clock,
38/211 D2.4: Impact of social media, Version V1, PU four hours after the initial blast, a lot of people were posting messages on Twitter and Facebook about their luggage and cars left at the airport: ‘This was growing and growing. So we knew that we have to make a decision about people luggage and cars’. However, because Coosto is only used if there is a huge crisis, staff are not as experienced at using it as those in Antwerp, Eindhoven Municipality or Rotterdam Safety Authority who use it on a daily basis and have learned to find the most relevant information.
What are the strengths and weaknesses of these tools? The choice of tools to use by the case study organisations appeared to depend on a variety of factors, including price, relevance and usability. Several of the respondents noted that they had changed the tools they used over the years or were considering doing so. Antwerp Fire Service was using Coosto but encountered some limitations in its ability to identify relevant information on social media in a timely fashion. One of the main reasons for this was thought to be the fact that Coosto has been mainly developed for commercial use – to monitor customers’ views and interactions about brands – and has not been developed with emergency services in mind. The Fire Service was therefore considering using a different, more expensive, social media analysis tool (PublicSonar).This tool was thought to have better search algorithms, as well as a database of local street names, and provides social media messages without a delay (for Coosto there is a delay of a few minutes after a message is posted) and has been developed specifically for emergency services – however, they were concerned about the increased cost of this service. Eindhoven Municipality selected Coosto following a sales pitch by three rival suppliers. Although the decision to choose Coosto was based on its technical functionalities and cost, an added bonus was that Coosto was developed in Eindhoven and its offices are in the city. Eindhoven Department of Communication report that they are ‘very happy with it’ and have experienced few problems using it. The only major problem cited is that sometimes the search terms used do not produce the expected results. This means that the analysts who use Coosto to monitor and analyse social media information have to be highly skilled and trained in their task. Eindhoven Municipality also benefits from direct support provided by Coosto in rectifying any issues encountered. The Netherlands East Region Police Service Centre trialled the use of Twitcident (which is now also known as CrowdSense or PublicSonar) as part of an annual fund-raising event called ‘The Glass House’. This is an event in December every year involving famous DJs to raise money for good causes. Twitcident was used to find out what was going on around a big event like that to monitor social media mainly for sentiment: ‘To try to find out how people are feeling; how many people are gathering and where; what are they doing and saying?’ This was to see whether it made sense to use a tool like this as an early warning system before something happens. This could include for example identifying Tweets about issues with crowds: ‘If someone sends a Tweet that it is very crowded where he is, but everyone is happy and cheerful, then there is no need to do anything. But if someone says: “it’s crowded but I don’t like the atmosphere around here now”, then I would suggest sending people around there to check it out and de-escalate the situation before anything more serious happens’. However, it was concluded that the needs of the service centre were sufficiently met through the use of a free tools such as TweetDeck or Hootsuite.
39/211 D2.4: Impact of social media, Version V1, PU Organisational structures and facilitators
How is the use and analysis of social media integrated into the emergency service’s organisational structures and systems? In most case studies, the use of social media is integrated into the emergency management cycle and that use of social media is an integral / routine part of their work and is supported by senior management. In Manchester, for instance, social media response is taken as seriously as a response to other approaches: “If you don’t deal with it, it makes you look bad.” In some but not all organisations this is guided by a social media strategy. Almost all case studies had staff with a role relating to social media, though it is comparatively rare to find roles that deal exclusively with managing and monitoring social media. Social media roles tend to sit in communications and / or public information departments or ‘digital services’ divisions and tend to be occupied by people with relevant (academic or professional) training (e.g. PR, communications), professional background (e.g. having previously worked for a social media company), or interest and ‘affinity’ (e.g. using social media and understanding the ‘language’). In Eindhoven, social media monitoring is carried out by a web care team put together via an internal collaboration between Communication and the Customer Contact Centre. Where the case study organisation is part of a multi-agency response structure, social media information tends to be shared across organisational boundaries. For instance, in Rotterdam- Rijnmond Safety Authority, social media strategy and use is coordinated by the Crisis Management Team, which involves representatives of a multi-stakeholder partnership involving sixteen municipalities, emergency services, regional government, other public entities and commercial organisations. In some cases, this extends to also involve citizens. In Rotterdam, for instance, emergency services use citizens as a resource via social media to add value to the information the services have, and fill in gaps. More broadly, our case studies reveal that the way an emergency organisation thinks culturally about social media and their role in emergency management influences how it is embedded in the organisations’ emergency management processes / procedures. From our case studies we can identify four organisational attitudes towards using social media in emergencies which are, arguably, themselves influenced by the wider local, regional if not national approach towards government and social media: The whole system enthusiasts (Antwerp, Eindhoven, Rotterdam) – using social media against the backdrop of a wider open government if not ‘smart cities’ background, these organisations are leading the way in demonstrating a confident whole system use of social media that benefits emergency management and includes citizens as an integral part of the social media strategy. The bottom up enthusiasts (Manchester/Netherlands East Region). Enthusiastically embracing the possibilities of social media, against the strong and confident backing of senior management these organisations significantly decentralize use of social media to local teams / staff. The top down professionals (Frankfurt, Brussels). Use of social media is centralised in a professional communications department and aided by a systematic accumulation of followers via different social media channels.
40/211 D2.4: Impact of social media, Version V1, PU The worried pragmatists (Graz). Recognizing the need to engage in social media for emergency response, this organisation has not yet worked through how to manage perceived risks and how to resource social media use for emergency management. The section below explores how embeddedness influences use of social media.
Which organisational structures or arrangements are likely to facilitate or hinder the use of social media in emergencies? Our case studies have revealed a number of enablers and barriers to use of social media in emergencies. These can be broken down into individual, organisational and systemic factors. Key factors identified by our case study work are listed in Table 3 below. In terms of enablers, it would appear that there are two key enablers at organisational level for social media use: Senior support in the organisation (management and / or board) has got a number of enabling functions for social media use. In Antwerp, for instance, it has helped expand its use in emergencies and because senior staff receive social media information it is taken more seriously by operational teams. In the Manchester case, the board and chief officer levels receive quarterly reports on social media analytics which measure engagement and analyse campaigns. An open and transparent organisational culture and one that constructively engages with potential ‘risks’ or ‘mistakes’ in social media use as learning opportunities supports the use of social media in emergencies. In Manchester, for instance, where social media use is decentralised down to the local police offers, there’s organisational acceptance that this devolution and the choice of a casual tone for social media communication will occasionally cause conflicts. Two examples and how the organisation deals with those are included in the vignette below.
For 90% of the content post there are no problems, but 10% will get a negative response (e.g. when the tone of a tweet is regarded as too light by media or the public) and the Greater Manchester Police board accept that as inevitable. GMP have made a conscious choice to adapt a light approach with their social media communication and the organisation has had bad publicity when the tone of a tweet is too light. For instance, one officer tweeted lots of song lyrics on one of the official Police accounts. The Manchester Evening News contacted the GMP Press team to say the officer had been doing this and they would run a small story about it. In those cases, the officer would be relieved of their tweeting duties, but as the officer in question was doing a tweet chat at the time, he tweeted publicly to say he had been ‘told off’ and was going off Twitter. Because of the public reprimand, the story got into the Daily Mail and became a big story. In another case there was a Yanks festival, and the traffic account tweeted a Nazi uniform at the festival. “Following this we had a very negative reaction from the Jewish community and we had to officially apologise for that”. These issues are seen as part of the social media model used by the GMP: “You’ve got to be prepared that we will have problems and that’s a risk that not every PCC is going to accept.”
41/211 D2.4: Impact of social media, Version V1, PU In addition to these two factors, having a sufficient number of suitably trained or knowledgeable staff commensurate with the (explicit or implicit social media strategy) is necessary. Social media use in emergencies can also increase the need for communication between different ‘arms’ of the response team. One example is offered in the vignette below. An example of difficulties that can arise between the social media team and other members of the service was provided during a recent industrial fire. Once again information about it was shared on Twitter within minutes of it starting: “There was a big industrial fire and the dispatching centre were upset because our communications people started tweeting about it before they’d had any calls about it – e.g. ‘1km away from the fire, close your windows’ – so people started calling into the dispatch centre from all around the city, asking: ‘We just saw your tweet. Do we need to close our windows too?’ But they didn’t have this information yet, we were ahead of the game and so they were upset about it. But I told them: ‘This just shows that you need to monitor social media’. But it also means we need to communicate better with them”. A tradition of openness and (technological) innovation in the organisation, city or region creates a (more) conducive environment for progressive and extensive social media usage across a whole emergency management system and beyond to also involve citizens and journalists. The vignette below illustrates for the case of Eindhoven how such systemic factors can shape social media use. Eindhoven Municipality has been a leading exponent of the ‘smart city’ movement and was an early pioneer in finding ways of applying information – especially ‘big data’ – to solve a range of problems, from traffic grid-lock, through developing the ‘circular economy’, to smart policing. Recent thinking around the ‘smart city’ has seen a shift in focus - from developing technologies that impose ‘order from above’ on citizens to create the perfect urban environment - to engaging with citizens as ‘co-producers’ of the knowledge required to make cities become the setting for a smart society. This has led Eindhoven to explore how social media can make a contribution. In line with other Dutch cities Eindhoven’s interest in social media reflects a broader objective of working with ways of delivering ‘open government’ in which all citizens are engaged as co-producers of information and knowledge. This means that 95% of the social media information processed comes from the general public. However, the key ‘influencers and movers’ within the community are well- known and in certain situations their activities on social media are monitored individually. For example, there was a recent situation in Eindhoven in which around 700 refugees needed to be accommodated. Social media was used in this situation to set up local area committees, with the support of key influencers and players, to handle the situation. Several case study organisations reported collaborating with other actors in the system (other first responders or public organisations, journalists, citizens) as part of their use of social media in the emergency management cycle: Rotterdam-Rinjmond Safety Authority has focused on developing effective collaborative networks with citizens as a whole; the Antwerp police service uses the same tool (Coosto) and shares information with them using a shared platform (called Telegram); Antwerp also recently started an experiment to look at new ways of communicating with journalists via social media during emergencies – so they are given access to a WhatsApp group and they can communicate with the fire service via
42/211 D2.4: Impact of social media, Version V1, PU that group; in Eindhoven, a multi-agency situation team (convened in emergencies) considers social media information to inform response. Table 3: Enablers and barriers to social media use in ES organisations
Enablers Barriers
Individual Affinity with or expert knowledge of Resistance to use social media (info) by social media frontline or dispatch staff: use of info not seen as valid or useful, staff feel
uncomfortable with the medium, staff Usage / strategy appropriate skills set do not sufficiently understand world of (soft, hard or both) social media.
Organisational Support for social media use by senior Organisational concerns about loss of staff and / or board of the organisation. control and reputation
Strong hierarchical structures can Insufficient staff for social media enforce compliance among reluctant ambitions. Other duties mean social staff with social media use / strategy. media accounts cannot be serviced 24/7. Difficult to keep staff well trained Organisational openness – to not considering pace of technological knowing or making mistakes. development Sufficient levels of suitably trained Fast pace of technological change staff commensurate with social media makes it difficult to keep staff well strategy (i.e. activities beyond trained. broadcasting will need more staff). Variable social media expertise Cross-hierarchy and / or cross- between departments and agencies departmental communication. can create inconsistencies of data analysis.
Systemic A tradition of openness and A conservative / traditional (technological) innovation in the organisational or city / country culture organisation, city or region creates a slows down extensive use of social (more) conducive environment for media in emergencies. progressive and extensive social media In multi-agency structures hierarchical usage across a whole emergency approval / decision making can slow management system and beyond to down fast reaction to social media also involve citizens and journalists. intelligence. Multi-agency response structures into
which social media use can be embedded.
Information sharing with other first responders.
43/211 D2.4: Impact of social media, Version V1, PU What ‘resistance to change’ dynamics or barriers in relation to the use of social media can be identified and how might these be addressed? On the whole, the case studies show relatively modest organisational resistance to the incorporation of social media into emergency service systems and practices. One organisational factor that impacts the scale of social media use is a concern about reputational risk and a lack of a strategy on how to deal with this. Lack of sufficient staff time dedicated to social media (e.g. because social media is part of a wider role) can limit the scale of social media use as accounts cannot be staffed 24/7. Variable social media expertise between departments and agencies can create inconsistencies of data analysis and hence reduces the potential operational use of the information. Several of our cases reported resistance to social media use by frontline or dispatch staff. Stated reasons for resistance included: staff not seeing the information provided by social media as valid or useful, feeling uncomfortable with the medium, or not sufficiently understanding the world of social media. In some cases, though, this has been effectively addressed through awareness-raising campaigns, education and training programs and, crucially, demonstrating to front-line staff that social media gets results. For example, in Rotterdam, live demonstrations to emergency services staff in control rooms of how social media information is used showed how it worked. In Manchester, police officers on the ground were initially resistant to using social media. Senior staff arranged for a journalist to work with the officers on a particular case – a rape case. The officers saw that the journalist was getting access to a wealth of information – particularly from the public – through Twitter and, eventually, more officers recognised the value of social media and began using it in their everyday work. The main barriers to the successful implementation of social media in emergency services are: The cost of commercial social media analysis tools Difficulties in using the analysis tools in particular in relation to choosing the right search terms to identify relevant information Some emergency services staff not fully understanding the ‘world’ of social media, the language used in it and how it operates A residual organisational culture in emergency services that devalues the contribution of ‘non-professionals’ Difficulties in keeping in touch with and responding to developments in social media technologies, their applications and their users Variability in expertise and experience across agencies and departments, leading to some inconsistency in the quality and usefulness of the social media analysis carried out, and highlighting the need for greater standardization in recruitment and training and the application of some kind of quality assurance system for social media analysis.
Conclusion An organisational perspective on social media use in emergencies might suggest that there are typologies and possibly even a trajectory of ‘sophistication of use’ (see Figure 19 below).
44/211 D2.4: Impact of social media, Version V1, PU One of the implications of this is that guidelines or ‘best practice’ guides need to speak to these different audiences to be useful and usable.
Top down ? professionals whole system Bottom up enthusiasts enthusiasts
worried pragmatists
Figure 19: A trajectory of social media use
Social media users and uses The case study analysis shows that emergency services and similar agencies typically use specialist social media teams embedded within emergency services to monitor and analyse information acquired through social media. In Antwerp, for example, the Fire Service has permanent staff with key responsibility for gathering and sharing information on social media. Any relevant information is shared with others in the organisation to provide situational awareness. In Manchester, the Digital Team within the police force – which is primarily responsible for managing social media - operates on a daily basis in the same control room with four other departments: Internal Communications, the Press team, Public Relations and the Awards team. Many of the functions covered by each team overlap, particularly those of the ‘Digital Team’ which is involved in proactive campaigns (PR), staff events (Internal Communications), reactions and appeals for criminals (Press) and long- service recognition (Awards). In one case – Frankfurt Fire Department – press spokespersons act as the main gatekeepers for dissemination of social media information. In addition to these ‘centralised’ specialist teams, social media is also used at grass roots level – but this varies from case to case. In Rotterdam, fire stations in the region have set up their own Twitter account and, currently, 20 out of 50 fire stations use it. In Manchester, 6000 police officers, and 350 people have Twitter accounts, of which 295 are active, and there are also 20 separate corporate accounts. As noted above, in Frankfurt Fire Department, only press spokespersons have access to organisational social media accounts and therefore are able to use social media over representative, official channels. However, employees especially from the voluntary fire departments can contribute by using their private social media accounts to actively send potentially relevant information or by passively being monitored by press spokespersons. The case study analysis further suggests that social media has become a catalyst for more wide-ranging and far-reaching changes to how emergency services interact with other
45/211 D2.4: Impact of social media, Version V1, PU agencies and stakeholders to deliver comprehensive and ‘joined up’ public services, not only through emergency services but more widely, to support public safety, and in the broader areas of communication, governance and accountability. In Brussels, the Crisis Centre provides services to the Federal Government and local municipalities in order to respond to crises. It has seven full-time staff with key responsibility for gathering and sharing information on social media during a crisis. It can also draw on the support of a wider team of experts (Team D5) based in local municipalities – who can provide support remotely during a crisis. In Eindhoven, routine monitoring of social media information on a daily basis is co-ordinated by a small team of 3-5 communication officers from the Department of Communication, who work closely with emergency services, statutory agencies and local journalists. All of these stakeholders are represented in emergencies via the ‘situation team’ that is convened when a crisis occurs. In Rotterdam, the Safety Authority has been set up as a multi-stakeholder collaborative structure, involving 16 municipalities in the Rijnmond area; emergency services; other public agencies and commercial organisations, including energy companies and 60 companies in the harbour and industrial areas - in cooperation with the municipality of Rotterdam. The Authority co-ordinates a General Emergency Control Room for Police, Fire brigade and Ambulance Services (112). In ‘crisis mode’, a ‘crisis management team’ from within the Safety Region will co-ordinate the response to an emergency. The Safety Authority incorporates a team of social media specialists to regularly monitor and analyse communication as well as provide key inputs in crisis management. In both Rotterdam and Eindhoven, social media has been used as an important tool by the municipal authorities in a broader agenda of establishing a culture of ‘information open- ness’. This has in turn laid the foundations for establishing the municipalities, and their public safety and emergency services, as credible and trusted sources of information for citizens and other stakeholders and has consequently increased the motivation of citizens and other stakeholders to play an active role as information providers and validators. A consistent theme across all of the case studies was the role traditional media – and journalists – play in emergency situations. The case study analysis shows that integrating social media with traditional media – and in particular creating the conditions whereby social media experts work with journalists – has had a number of positive impacts. These include increasing the credibility of social media, supporting buy-in for social media at all levels (strategic, management and ‘grass roots’) and contributing to information validation and countering misinformation. In Manchester, for example, police officers on the ground were initially resistant to using social media. Senior staff arranged for a journalist to work with the officers on a particular case – a rape case. The officers saw that the journalist was getting access to a wealth of information – particularly from the public – through Twitter and, eventually, more officers recognised the value of social media and began using it in their everyday work. In Rotterdam, journalists are key consumers of information provided by the Safety Authority. They then pass on this information to the public and other stakeholders, for example by re-tweeting. Journalists are also important in the information validation process because they act as a point of triangulation for information collected and analysed by the authorities. In Rotterdam, information collected from journalists and other partners in the network is
46/211 D2.4: Impact of social media, Version V1, PU published on the Safety Authority’s live blog, after checking its validity. In Antwerp, the Fire Service recently started an experiment to look at new ways of communicating with journalists via social media during emergencies. Journalists are given access to a WhatsApp group and they can communicate with the AFS via that group. They also have agreements with newspapers to use AFS tweets in their communications. In Eindhoven, officers from the Department of Communications who are responsible for monitoring social media work closely with local journalists.
What types of information are emergency services most interested in? In three of the cases – Eindhoven, Rotterdam and Brussels – a similar social media information analysis model is being used to carry out what is referred to by the Rotterdam Safety Authority as ‘environmental analysis’. The model is based on an information ‘triad’ that attempts to capture and analyse three main types of social media information: i) Questions (how many victims are there? where is the situation happening? is there a lack of information? are people asking particular questions?) ii) Behaviour –what are people doing right now? are people moving towards the area where a disaster is happening and do they need to be warned?) iii) Emotions (what feelings and sentiments are people expressing about a situation? what are people most worried about? do they need to be reassured?). By interrogating the information coming in through social media according to these three types, social media analysts can make a distinction in an emergency situation between people who are actually affected; people who think they are affected and people who are not directly affected. This interrogation helps the filtering process because sometimes the system has to analyse many thousands of social media items. It can also help the team to decide on an appropriate response. Questions tend to suggest that people are aware of a situation but are not at the centre of it, and so need a response that is informative and reassuring. Behaviour tends to suggest that people are actually experiencing something directly, and is often signalled by the use of personal pronouns in a social media communication, for example ‘I’ve just seen an explosion in a factory’. Behaviour-type communications are typically associated with an ‘action’ response – for example sending a fire truck to the vicinity mentioned in the communication. Emotions tend to suggest that people have heard about something and are worried, and emotion-type communications typically require issuing of alerts, backed up with informative and reassuring messages. In general, emergency services are interested in any information that can help them carry out a situation assessment/environmental analysis and an assessment of the actual and potential impact of the emergency – especially in terms of gauging the ‘public mood’ of citizens. They are particularly interested in eyewitness accounts, especially when supported by posts of images showing the incident. For example, during a recent storm in Antwerp, social media was used to explore the impact of the storm and to compile damage reports. The fire brigade saw that people were sharing a picture of trying to put a tarpaulin on a damaged roof to protect a historical building. “So, we contacted them via social media and said: ‘Wait for us, we’ll do it. It is dangerous – please direct message us’. Then we swapped numbers and spoke to them directly to arrange a resolution”.
47/211 D2.4: Impact of social media, Version V1, PU What kinds of social media citizen roles would be most useful for emergency services? None of the cases included a formal role for citizens within the systems and processes set up by emergency services and municipalities to analyse social media information. However, in all of the cases a strong emphasis was placed on engaging citizens as active information providers and validators. In Antwerp, the Fire Brigade has links with established citizen groups to enable them to use their support in emergencies: “We have contact with an established movement – Antwerp Speaks – a conglomerate of all citizen organisations.” In Manchester, Greater Manchester Police work closely with community mediators, but are less enthusiastic about ‘citizen journalists’, because they find that citizens do not understand the legal restrictions governing publication of information (for example publishing names of victims; transgressing contempt of court rules). In Brussels, the Crisis Centre does not specifically ask for information from citizens but draws on information shared between citizens on social media as well as any concerns directed at them via social media or the contact centre. More broadly – especially in the Netherlands – the role of citizens in social media analysis for emergencies can be seen as part of a much broader movement towards supporting the increased democratisation of local government and local politics. In Rotterdam – which has long been part of the ‘Smart Cities’ movement - the social media citizen role the Safety Authority wants to encourage is a ‘collaborative’ or ‘co-producer’ role. The emphasis in Rotterdam-Rinjmond is on trying to bring citizens as a whole into the multi-stakeholder partnership and network set up within the region to deliver safety. The ultimate aim is to make interaction between citizens and emergency services part of everyday life. Because the system is based on a holistic approach, engaging all stakeholders within the Safety Region – including citizens and journalists – as co-collaborators in creating, accessing and analysing social media information, the system operates as a ‘multiplier’ for information. Similarly, the city of Eindhoven has worked to develop an ‘open government’ environment in which all citizens are engaged as co-producers of information and knowledge. This means that 95% of the social media information processed comes from the general public. However, the key ‘influencers and movers’ within the community are all well-known and in certain situations their activities on social media are monitored individually. For example, there was a recent situation in Eindhoven in which around 700 refugees needed to be accommodated. Social media was used in this situation to set up local area committees, with the support of key citizen influencers and players, to handle the situation.
Information validation
How can emergency services be persuaded that information provided via social media from citizens is credible and trustworthy? The case study analysis showed that only Rotterdam-Rijnmond Safety Authority has invested in awareness-raising and promotion to reinforce the value the Safety Authority places on information provided via social media from the citizens and their credibility. One example is the demonstration of the real-time dashboard which persuaded the emergency service of its value. In Frankfurt Fire Department, there is no need to persuade emergency service that social media information is trustworthy due to the hierarchal structure and the press spokesman being in charge. There are guidelines about the use of social media and
48/211 D2.4: Impact of social media, Version V1, PU this is well integrated in their daily work. Eindhoven Municipality and Rotterdam-Rijnmond Safety Authority highlighted that they are open government agencies which helps to build trust and credibility among citizens. This will then in return motivate citizens to provide accurate information and to be willing to collaborate with the authority if rumours are encountered.
What procedures and tools are used to validate such information efficiently and effectively?
The analysis revealed that case studies differ in their view of false rumours and misinformation on social media being an issue or not. Eindhoven Municipality and Antwerp Fire Service reported that reliability of information on social media is not a severe issue for them and they were not aware of any recent examples. In contrast, Graz fire service struggled with the possibility of information being misleading and also Rotterdam-Rijnmond Safety Authority stated that false rumours are an ongoing issue for them. There was for example a false rumour about a gunman in a shopping mall spreading very quickly on several social media platforms four years ago. For the Brussels Crisis Centre rumours have been an issue in some incidents but not in others. During the Brussels bombing, information validation was not seen as an issue as the focus was on citizen’s concerns. However, in another recent terrorism related crisis rumours spread on social media and citizens were asked to stop sharing misleading photos.
Most case studies use data triangulation to decide if information on social media is credible and trustworthy. However, case study areas vary in the type of information they triangulate. Graz Emergency Service and Antwerp Fire Service stated they compare the respective information with information from other sources or other social media messages. Both case studies further judge the credibility of photos and videos by looking at potential errors in terms of location or time of the day. Eindhoven Municipality applies a two-step process of triangulation. First, the team of analysts compare information; following that members of the situation team (police, fire department and health agencies) cross-check information further, also with local media stories. Rotterdam-Rijnmond Safety Authority has an overall ‘environmental assessment’ in place. They check social media information with 112 information room emergency desk, with emergency staff operating on the ground at the scene of an incident, with journalists and with feedback from citizens. Only Greater Manchester stated that they use information about the account, e.g. numbers of followers, to decide about the trustworthiness of information.
Emergency service staff skills and resources
What skills are required to help emergency service staff to find, interpret and make use of social media data?
In relation to the staff skills required to make use of social media within emergency services, a strong set of themes emerged across the case studies. The skills outlined fall into two broad categories. The first of these can be termed ‘technical knowledge and skills’ and the second ‘soft knowledge and skills’.
49/211 D2.4: Impact of social media, Version V1, PU Interviewees pointed to the need for those staff directly involved in monitoring social media to be able to use and have a working knowledge of a range of social media platforms. Most notably these referred to Twitter and Facebook as these are both most popular and most amenable to emergency services; however newer platforms such as Instagram and Snapchat were also referenced. For each platform, ES staff need to understand the basic functionality, for example with Twitter knowing how to ‘@’ other users, how to ‘retweet’ messages or how to search keys words or ‘hashtags’. Equally, interviewees across the case studies spoke of the need to be skilled in the use of social media analysis tools if they are used within their service, the most common of which was Coosto. This particularly related to the ability to programme social media search strings. These were described as often long and complex, as they need to reflect: the range of possible emergencies that the service is responsible for; the local geography and dialects; and as far as possible to exclude false positives from being picked up.
In addition to the technical skills and knowledge needed by those most directly engaged in the monitoring of social media mentioned above, interviewees also pointed to softer skills and knowledge required to successfully use social media within their services. These included a general familiarity with social media platforms or what one ES staff member described as a “feel for the networks” (this applied to the wider team who may need to incorporate social media information into their operations). This means that staff need to understand how the platform is used by the public, how users interact with one another and how language is used on different platforms, such as common slang terms, abbreviations and acronyms. The need for familiarity also extends to the local area and culture, including area, street and building names and local slang and dialects. This familiarity supports the interpretation of social media data but also allows staff to interact with other users in a natural way when using social media. To this end, communication skills and an ability to ‘read between the lines’ were also mentioned as important skills for staff to possess, as were emotional intelligence and empathy. These attributes were seen as underpinning staff’s ability to interact with the public via social media platforms in an informal and conversational style and their ability to read the mood and sentiment of the community. Being empathic and diplomatic was also seen as helpful in communicating social media data to colleagues in operational roles. This means the ability to frame advice for the operational team in ways that will see it used. Other desirable facets included being fast and efficient in terms communication and having good problem solving skills. As one interviewee summarised: “The best social media analysts are those who like solving puzzles and people who can solve things through trial and error. Also, you have to think like the man in the street. You need to know their slang and have their local knowledge”.
What guidance or training would be most useful for them?
ES staff suggested that training that helps to instil the knowledge and skills described in Section 5.5.1 would be useful in improving their use of social media. At its most basic level staff called for technical training around the use of specific platforms and analytic tools. Suggestions also included training to understand how others use social media and their behaviour on these platforms to address the need for familiarity for social media platforms. However with some of the softer skills and the familiarity with platforms and social media culture one interviewee suggested that these skills are hard to teach: “You need a special type of person involved – we need people who are digital natives. Training people who are
50/211 D2.4: Impact of social media, Version V1, PU not is very hard” bringing into question the value of trying to upskill staff who are unfamiliar with social media. One interviewee raised the specific need to learn “socio-psychological strategies to counter irrational behaviour in social networks.”
A key need identified across many of the case study organisations was that any training or guidelines provided would need to keep up-to-date with the changing landscape of social media use. It was noted by many that social media platforms can experience rapid changes in popularity and use and so training would need to be regularly revised to reflect this or encompass the use of a wide range of existing platforms. Moreover, some interviewees suggested that there may be extra value in providing standardised training or guidelines for complementary agencies (for example to a municipality’s police, and ambulance and fire service). This would help address the lack of standard uses and approaches to social media between services. Interviewees also suggested simple guidelines that could be widely available and useful during emergencies.
Moderating citizen communities
Substantial variations were found between the different case study organisations in the way they interact with citizens, the importance they place on cultivating links with citizens and the roles that citizens take up in the emergency management cycle. This variation appears to depend on the Country, city, type of ES involved, and the history of development of social media use within the service. Despite this wide variation the case studies point to a trend of services moving away from working with particular citizen communities and instead putting resources into developing effective collaborative networks with citizens as a whole. However in a number of areas there as still a perceived benefit of cultivating links with existing community groups and prominent individuals to help gain and disseminate information at all stages of the emergency management cycle. More broadly all the services saw the value of using social media to communicate directly with citizens for the purpose of broadcasting key messages and helping to improve the public image of their services.
How important are public/volunteer social media communities in supporting emergency services?
Interacting with citizens and with volunteer communities via social media was seen as having a number of possible benefits for emergency services. Some organisations spoke of spontaneous volunteer groups emerging during major emergencies or events, such as the Refugee Aid drive in Frankfurt in 2015. The value of social media in these examples was the relative speed and spontaneity with which volunteer communities could form. These groups were seen as able to support emergency service efforts through the provision of specific skills, such as translation, and general support such as providing supervision. This point was supported by the Greater Manchester Police who spoke of receiving support in the identification and apprehension of suspects through social media as well as facilitating the clean-up and donations after riots. Despite these benefits some respondents suggested that during major events the process of organizing large numbers of volunteers could itself become resource intensive particularly at a moment when emergency service staff were most under pressure themselves so there was perceived to be a trade-off in terms of the advantages and disadvantages.
51/211 D2.4: Impact of social media, Version V1, PU Most services, but particularly the police force, saw social media as a way to improve public relations. As a respondent from the Greater Manchester Police (GMP) explained, the service has had a poor reputation and low levels of trust with the public. As part of a wider drive to address this, the service uses social media to build relationships with the community and present a more approachable face of the police. Social media was also seen as one way to feedback to the public about the impact volunteers have had thereby increasing the public perception of volunteering and helping to maintain their motivation.
How do emergency services engage with these communities? In the coordination of spontaneous communities such as those referred to in the previous section the service reported nominating a named contact person from within the volunteer community to communicate through and to improve coordination of volunteer efforts. Beyond this respondents spoke about the use of social media to broadcast advice and provide information to existing SM media groups and communities via the major platforms such as twitter and Facebook. In some cases services have standing relationships with existing community groups or high- profile individuals in the local area. In Frankfurt, for example, there are plans to create a Virtual Operation Support Team (VOST) of volunteers who can be activated in emergencies or when there is need. In Manchester the police service has links to existing community groups such as ‘community watch’, a neighbourhood safety initiative, and will both post to and read their Facebook group. In Eindhoven the social media activity of key ‘influencers and movers’ from the local area is monitored and sometimes they are brought into special committees set up to respond to a particular emergency situation. Other areas, such as Eindhoven and Rotterdam prefer to focus on the development of relationships with citizens as a whole, rather than cultivating citizen communities per se. Nevertheless a number of areas spoke of developing Innovative ways of engaging with the public around social media. In Manchester the police force run events with the public known as hackathons where members of the public are supported to develop new apps and Community reporters are also invited to ride along with police and tweet and post about their experiences. Rotterdam spoke of experimenting with Facebook groups for incident prevention and are currently, producing a ‘Peter the Fireman’ blog to raise awareness around domestic fires. A number of respondents were keen to note that while volunteers may be called on around a specific emergencies or event there is a need to maintain engagement at all times. As one respondent noted: “You can’t start doing social media when the streets are on fire”. For this reason many of the services try to maintain a regular social media presence. This was found, for example, in the Antwerp fire service, where content that is thought to be appealing to the public, such as behind-the-scenes tours and videos of pet rescues, are regularly posted on social media. These activities are both to improve public perception and also to cultivate a large follower base should the service need to broadcast warnings or advice.
52/211 D2.4: Impact of social media, Version V1, PU What should be done to support social media volunteer communities in emergencies to help them become more efficient and effective? Few suggestions were made in terms of ways to improve volunteer involvement via social media beyond developing better structures through which to coordinate volunteers during emergencies or in the case of Frankfurt the development of a Virtual Operations Support Team (VOST). One further suggestion was to undertake a skills audit of IT and social media skills among regular volunteers so the service is better aware of its capacity and better able to call on particular individuals when their skills may be needed.
What can emergency services do to make stronger links with such communities before, during and after emergencies? A key approach that many services pointed to was the cultivation of a significant social media presence and correspondingly large follower base. This then allows services to promote their messages at all points in the emergency management cycle – providing advice and warnings before an event, proving up to date information during an emergency and then supporting citizens in the aftermath. It was noted that to maintain a large follower base, content needed to be a mixture of information and public interest material. Beyond this, a number of services suggested possible ways to better engage citizens. For example Feuerwehr Frankfurt is planning to provide apps and create a Facebook groups so as to permit greater direct interaction with citizens during emergencies; the Antwerp Fire Service is considering the use of a bot on Facebook Messenger to answer simple questions during an emergency, therefore freeing up staff to concentrate on more urgent matters.
Lessons learned and implications for EmerGent There are now many rival tools available to analyse and manage social media data: this round of case studies has shown that there are now many different tools available to collect and analyse social media data. Some quite basic ones are available free of charge, but those providing more complex functionalities are available only at a cost to emergency services and other organisations. This is quite a different landscape from the start of the project (in 2014) when such tools were either in their infancy or had not yet been developed – the growth of social media and recognition of its importance has led to a rapid expansion of tools to harness such data. However, very few of these tools are developed with emergency services in mind – instead, most of them were developed mainly for commercial organisations – and there are still key areas that need further development, including: Geographical location of messages: social media analysis tools are usually not able to identify the exact location of social media messages sent – some tools have tried to address this by providing data-bases of local place names (streets, towns, etc.) but these only work if the Twitter or other social media messages use such ‘locations’ False positives: despite the use of complex search strings and other functions, most tools still provide a lot of information that is not relevant to a particular emergency or service; this means that during a major incident those using such a tool may struggle to identify relevant messages or resort to searching by hashtags (#) only which can result in missing other important information not shared in this way
53/211 D2.4: Impact of social media, Version V1, PU The cost of such tools: is a barrier for some emergency services with limited budgets – this suggests a need for a free or very low cost option that goes beyond the very basic functionalities available in free tools such as Hootsuite or Tweetdeck Types of social media data: most of the tools available at the moment focus on collecting information from Twitter and/or Facebook only – given the growth of alternative, less open, social media channels such as Snapchat, Instagram or GoBubble, there is a need for tools to consider ways of accessing information from such channels as well. Social media can inform the emergency services response in different ways: the case studies of ‘high end’ ES social media users showed that such data can provide different types of information depending on need. In particular, this included three types of information: a) situational awareness: How many victims are there? Where is the situation happening? b) behaviour: What are people doing right now? Are people moving towards the area where a disaster is happening and do they need to be warned? c) emotions: What feelings and sentiments are people expressing about a situation? What are people most worried about? Do they need to be reassured? This means that: Social media data can provide a valuable way of collecting additional information to supplement data received during emergencies in the customary way (112 calls, and information collected first hand by ES staff) to help ES respond in the best possible way Such information can inform the messages sent out by ES during an emergency. Thus, for example, questions sent via social media by citizens tend to suggest that people are aware of a situation but are not at the centre of it, and so need a response that is informative and reassuring. Messages describing behaviour tend to suggest that people are actually experiencing something directly, and need advice or support on how to keep safe. Emotions tend to suggest that people have heard about something and are worried, and require issuing of alerts, backed up with informative and reassuring messages. Training in using and analysing social media needs to be directed at right staff: One of the key findings of Deliverable D2.3 was that lack of organisational structures, procedures and staff skills were some of the key obstacles to the use of social media data by emergency services. This round of case studies has explored this further and shown that staff tasked with analysing social media need to be supported by other organisational structures and systems, and have both the technical and soft skills to understand and engage with social media data. This means that it was very challenging to upskill those who are not ‘digital natives’ – who are not familiar with the use and ‘syntax’ of social media and / or lack the emotional intelligence and empathy to engage with the public via social media platforms in an informal and conversational style. Guidelines on using social media need to be tailored to existing level of social media use: analysis of the use of social media in case study organisations suggested a trajectory of use ranging from ‘worried pragmatists’ to ‘whole system enthusiasts’ – with very different ways of engaging with and using social media in their organisations. One of the implications of this is that guidelines or ‘best practice’ guides need to speak to these different audiences to be useful and usable.
54/211 D2.4: Impact of social media, Version V1, PU 6 Use of social media in response to terrorism
The Brussels bombings The purpose of this case study is to explore the way social media is used during emergencies and in particular how social media use interacts with local emergency responses in the event of a terrorist attack. These questions are explored in relation to the terrorist bombings in Brussels in March 2016.
Location and boundaries of the emergency The City of Brussels is the capital city of Belgium with a population of over 176,000 people [WWW11]. It sits within, and is the largest municipality of the wider metropolitan area Brussels-Capital Region which has a population of around 1.2 million people. Along with being the capital of Belgium, Brussels is the main seat of the European Union (EU) Institutions and houses the EU commission and the European Parliament. As a result, a large part of the Brussels population is not Belgian with an estimated 30% of the population born outside the country [WWW12]. The City of Brussels is located near the centre of the country as Figure 20 shows. Figure 20: Location of Brussels within Belgium
Source – Google Maps
Main events and key incidents On the morning of the 22nd March, 2016, three coordinated suicide bombs were detonated in Brussels. Two bombs were set off in Brussels Airport in Zaventem and one at the Maalbeek Metro Station. The first explosion happened at 7:58 am at the Airport and was followed
55/211 D2.4: Impact of social media, Version V1, PU seconds later by a second explosion.[WWW13] Investigation in the immediate aftermath of the event showed that there had been a third bomb in the airport, but the force of the first two explosions had prevented it from being set off. A third bomb was successfully detonated a little over an hour later at 9:15 am, six miles from the airport on a train leaving Maalbeek Metro station, near the European Commission headquarters (See Figure 21 for locations of the two bombing sites).
Figure 21: Sites of the Brussels bombing of 2016.
Source: Google Maps
The BBC described the events on that day in the following way [WWW14]: “Two explosions, moments apart, tore through the check-in area of Zaventem airport at 07:58 local time (06:58 GMT). One eyewitness reported hearing shouts in Arabic just before the explosions. Officials said the bombs were detonated just seconds apart at opposite ends of the departures hall. Witnesses said people ran from the site of the first blast, only to be caught in the second, near the main entrance. Just over an hour later, a further blast happened at the Maelbeek metro station in the city centre, close to several EU institutions. A three-carriage train was just leaving the station in the direction of Arts-Loi, the next stop a short distance away, when the blast happened, according to Brussels transport operator STIB. The bomb was apparently detonated in the middle carriage, which was running along the platform at the time. The driver immediately halted the train and evacuated the carriages”.
56/211 D2.4: Impact of social media, Version V1, PU Figure 22: Location of the two explosions at Brussels Airport
Shortly after the attacks, a number of precautions were taken including the evacuation of the airport and stopping of all air traffic in and out of Zaventem. All public transport was stopped and key European Commission buildings were evacuated or given extra security. Police also evacuated schools and universities, and ordered workers in surrounding office buildings to remain inside. The national and European Union terror levels were raised to their highest level (‘four’ and ‘Orange’ respectively) as a search for people involved in the attacks continued. Soon after the attacks ISIS released a statement taking responsibility entitled “Statement on the Blessed Raid in Brussels against Crusader Belgium.” [ICT16] In response to the attacks police made raids on premises in Brussels after receiving tip-offs from the public, finding explosives, radical Islamist paraphernalia and a laptop containing a suicide note from one of the bombers [WWW15]. Multiple arrests followed, with to-date twelve men having been arrested in relation to the attacks [WWW16]. The attacks represented the worst terrorist acts in Belgium’s history [NeSt17] although Belgium has been subject to a high number of attacks over the last two decade [ICT16]. Thirty-two members of the public died in the attacks as well as the three suicide bombers. Over three-hundred others were injured. Around half of the fatalities were Belgian citizens with others largely coming from European countries and America [WWW17].
The role of the National Crisis Centre The Belgium National Crisis Centre of the Federal Public Service (FPS) Internal Affairs provides its services to the Federal Government and local municipalities in dealing with and responding to crises. One of its main tasks is to collect, analyse and disseminate the necessary information to the political and executive authorities. The Crisis Centre also offers its infrastructure and know-how in interdepartmental management and crisis coordination at the national level. During a crisis, the Crisis Centre will provide advice on what, when and
57/211 D2.4: Impact of social media, Version V1, PU how to communicate information to the public and discuss the overall communication strategy with the crisis team. The communication service of the Crisis Centre employs seven staff full-time (up from four staff in March 2016) and can also draw on the support of a wider team of experts (Team D5) based in local municipalities – who can provide support for crisis communication. The Crisis Centre can also draw on the services of a contact call-centre which citizens can call to share and receive information during a crisis. On 22nd March 2016, the Crisis Centre was hosting a planned crisis communication exercise with the support Team D5 – which meant that 25 members of the Team D5 were on the way to the Brussels head-office at 8am, just as the attacks at the airport started on that day.
Patterns of social media usage The attacks prompted a large volume of social media traffic and the hours and days following the events saw over 2 million messages posted on Twitter on the subject.17 In keeping with the nature of social media use, much of this traffic came from outside Belgium showing that the response to the attacks in Brussels was global instead of just national or regional. Figure 23 shows the global density of twitter traffic related to the attacks in the first 24 hours after the attacks took place [BDMR16].
Figure 23: Global Density of Twitter Traffic – Brussels bombings During and in the immediate aftermath of the bombings the Belgium Crisis Centre released statements through all outlets including Twitter18 that perpetrators may remain at large and encouraged the public to use social media or text messages to contact friends and family to prevent a surge in demand leading to the saturation of the telecoms networks [Pete16]. Several other organisations also took steps to support the emergency response to the
17 Organisational case study, section 4.5 18 https://twitter.com/CrisiscenterBE/status/712207628571643905
58/211 D2.4: Impact of social media, Version V1, PU incidents including the major provider of Wi-Fi services nationally, Telnet Hotspot, opening up its service free of charge.19 Three hours after the airport bombs, Facebook made available its Safety Check function allowing people in Brussels to indicate to their friends that they were safe, thereby reducing strain on the telecoms network [Alap16]. Google followed suit by providing a link to the Crisis Centre website below their search box and Twitter offered support to the National Crisis Centre who asked the platform to put a ‘push’ message to all Twitter users currently in Belgium telling them to look at the Crisis Centre website for official information about the attacks.20 Hashtags expressing solidarity with the victims of the attacks and with the city and country as a whole started to trend on Twitter around the world. The most prominent expression was the French-language hashtag ‘#JeSuisBruxelles’ (‘I am Brussels’) echoing the slogan ‘#JeSuisCharlie’ coined after the attacks on the Charlie Hebdo offices in Paris the year before. Images of Tintin, the Belgian comic character, crying and images of the Belgian flag were also shared widely on Twitter and Facebook as symbols of solidarity [ICT16]. At the same time hashtags such as #ikwilhelpen (‘I want to help’) and #PorteOuverte (‘Open door’) were posted by people in and around Brussels, offering places to stay for those unable to travel due to the attacks and assistance for people who might have been affected by the attacks. At the same time groups and individuals supportive of the attacks also used social media to promote their messages. Known Jihadi voices on social media made supportive declarations and encouraged others to continue with terrorist activities in Europe. These conversations also used a range of hashtags such as “Brussels is burning”, “The Explosions of Brussels” as well as others [ICT16]. Moreover, some actors supportive of the attacks posted misleading or outright false information including videos and pictures taken from the scene of the bombings, some of which were unclear or faked images of other attacks. In the most extreme cases, Jihadists’ accounts broadcast fake information including the claim that bombs had also been placed in Brussels Free University and St. Peter Hospital in Brussels in an attempt to further scare the public and interrupt the emergency services response to the attacks [ICT16]. The role and use of Twitter is of particular importance in emergencies as it is open and therefore allows messages to be disseminated widely and at speed without direct connection between source and eventual recipient. A study undertaken by the European Emergency Number Association (EENA) attempted to track the various stages of Twitter response to an emergency, drawing partly on data from the Brussels bombings [Vand16]. The paper argues that there are six distinct phases to an emergency as seen by the different ways people behave on the social media site and the different actors that are engaged at different times. Stage 1: information Phase sees users learning about the event and as such tweets and responses are often characterised by their informational content and ‘neutrality’ of tone’. (Figure 24 shows examples of this kind of tweet). Stage 2: Phase of emotion: Following on from this, as the nature of the event becomes clearer, individuals on twitter begin to express emotional responses to the incident. These tend to display surprise, shock, sadness and anger (Figure 25). These are followed by Stage 3: Transitional phase, which
19 https://twitter.com/CrisiscenterBE/status/712219220579319808 20 Organisational case study, section 4.5
59/211 D2.4: Impact of social media, Version V1, PU may occur at different times for different groups of users, and this phase characterises late comers to the event combining the two previous phases in one. Stage 4: Organisational phase sees a consensus start to emerge online about the nature of the event and the right responses with discordant voices being pointed out and criticised, this phase also sees a migration away from informational hashtags characterised by place names descriptions of the event towards hashtags expressing support and solidarity such “#PrayForBelgium” or images of the Belgian flag. Stage 5: Phase of interest, the term ‘interest’ is used in both its meanings. This sees some online actors promoting their brands through the medium of expressions of solidarity but also this phase sees greater detail of the events being released by state and media actors. Stage 6: Disorganisation phase: Sees a splintering of groups away from the mainstream hashtags, often these divisions occur around differing politics, that find groups dissenting from the main response, disavowing the main hashtags, for example people not wishing to “Pray for Belgium” as this is too religious. Figure 24: Tweets from the information phase
Figure 25: Typical tweet of phase of emotion
Detailed Analysis of Twitter: scale and sentiment of tweets The Tavistock Institute also undertook primary research activities with social media data (see Section 2.4) to try to understand the ways Twitter was used in the aftermath of the bombings. Social Media data was collected during the Brussels attacks in March 2016. Analysing social media data can help to understand how social media is used during emergencies in many
60/211 D2.4: Impact of social media, Version V1, PU ways, including who uses it, when it is used, why it is used and what content is embodied in the social media posts. Here, we present the analysis of Twitter data with regards to the scale of information and sentiments of tweets. The latter served more as an explorative analysis using machine learning algorithms. As the first dataset was substantially smaller we decided to mainly use the second dataset for the analysis. The second data set consisted of 2,857,593 posts on Social Media during the timeframe 11.44am on the 22nd March and 9.05pm on the 23rd March 2016. The actual attacks however, happened earlier that day, more precisely at 07.58am at Zaventem international airport and 9.11am at Maelbeek metro station in the city centre meaning that we do not hold data from the first few hours following the attacks. As stated above the dataset did not only contain Tweets but also posts from other Social Media platforms. The vast majority of messages about the attacks were posted on Twitter (see Table 4). As the nature of messages on the various platforms is very different from each other, we decided to only include messages from Twitter in the analysis.
Table 4: Number of messages per platform Social media N % platform
Twitter 2791271 97.68%
Instagram 49163 1.72%
Google+ 12002 0.42%
YouTube 5136 0.18%
Facebook 21 0.00%
The data set also did not only consist of English tweets but also ones in many other languages. The condition was only that either the term Brussels or Bruxelles was included in the post. See Figure 26 for a breakdown by language; only languages that were represented in at least 1% of the Tweets are displayed in the figure. We only selected English messages for the more detailed analysis as we further aimed to explore the sentiments expressed in the tweets. Finally, restricting the dataset to messages on Twitter only and in English resulted in 2,234,850 tweets.
61/211 D2.4: Impact of social media, Version V1, PU Figure 26: Percentage of posts by language
90% 79% 80% 70% 60% 50% 40% 30% 20% 10% 10% 4% 2% 1% 1% 1% 1% 1% 1% 0%
Figure 27 illustrates the development of the volume of tweets per hour on the 22nd and 23rd of March from 11am onwards. The highest number of tweets was collected between 6pm and 7pm CET with 137,688 tweets. After 11pm there is a significant decrease from 100,147 to 66,750 tweets. There are tweets missing between 4pm and 5pm. We supplemented tweets from the first dataset harvested with R; even with the additional tweets there seems to be a gap as apparent from the graph below.
Figure 27: Number of tweets per hour.
160000
140000
120000
100000
80000
60000
40000
20000
0
1am-2am 2am-3am 3am-4am 4am-5am 5am-6am 6am-7am 7am-8am 8am-9am
1pm-2pm 2pm-3pm 3pm-4pm 5pm-6pm 6pm-7pm 7pm-8pm 8pm-9pm 1pm-2pm 2pm-3pm 3pm-4pm 4pm-5pm 5pm-6pm 6pm-7pm 7pm-8pm 8pm-9pm
9am-10am
11am-1pm
4pm-5pm*
12pm-1pm 9pm-10pm 9pm-10pm
12am-1am
10am-11am
11am-12pm 11pm-12qm 10pm-11pm
62/211 D2.4: Impact of social media, Version V1, PU Among these tweets there are a high number of retweets. The exact number is not available for this dataset; for the first dataset it was found that 83% are retweets and only 17% original messages. The highest retweet count is 148,568. The main aim of the analysis was to explore the sentiments expressed in tweets and the change of sentiments over time. Due to the high number of tweets, manually exploring the nature of the tweets was not possible. To ease the process machine learning algorithms were applied. First of all, tweets were prepared which included deletion of punctuations, html links, @people, numbers and the term ‘RT’. This was done using R. As we wanted to study the emotions expressed in the tweets we needed a classification of emotions. Often, sentiment analysis is based on positive, negative and neutral emotions. However, for the analysis of tweets related to the Brussels attack we sought to be more precise and searched for other classifications. We decided on the Plutchik's wheel of emotions which consists of eight emotions [MoTu13]. The Plutchik's wheel of emotions is an extension of Ekman 6 emotions which additionally includes Trust and Anticipation. The list of eight emotions are: Sadness (grief, pensiveness) Fear (terror, apprehension) Anger (rage, annoyance) Disgust (Loathing, boredom) Surprise (amazement, distraction) Anticipation (vigilance, interest) Trust (admiration, acceptance) Joy (ecstasy, serenity) Emotions in brackets represent related emotions to the eight main emotions. We then randomly selected 1000 tweets to be coded. During this process we realised that there was a need for one additional category that comprises news, information and updates that did not contain any emotion. We referred to this category as ‘Neutral’. Furthermore, there were no tweets that expressed ‘Joy’ so this emotion was omitted. The emotion ‘Anticipation’ was found very difficult to code and was only found in a very few tweets and was for this reason excluded from the sentiment analysis. After the first coding process 6 sentiments and the neutral category remained. It should be noted that tweets are in general challenging to code due to their style and structure. Tweets are very short messages, and sometimes two emotions are expressed in one message or the sentiment is hidden or ambiguous. Since some emotions were less frequent than others we coded a number of additional tweets to reach a reasonable sized trained set per emotion. The set of coded tweets was divided into training and testing datasets. Previous work that we have conducted has shown that commonly used machine learning algorithms perform best when dealing with tweets. We applied a variety of machine learning algorithms, namely Support Vector Machines (SVM), Maximum Entropy classifier (Maxent), Random Forest (RF), and Bootstrap aggregating (Bagging). We used approximately 1/20 of the total amount of tweets for testing with the remainder used to train the algorithms [WWW18]. Training the algorithm was a continuous process by providing feedback to the system and feeding it back to the training dataset. The last round achieved prediction accuracies of around 90% as shown in Table 5.
63/211 D2.4: Impact of social media, Version V1, PU Table 5: Prediction Accuracy per machine learning algorithm Algorithm Random Maxent Bagging SVM Forest
Prediction .86 .90 .80 .90 Accuracy
Maxent and SVM performed equally well; as we have applied Maxent algorithm in previous work we adopted Maxent to predict the sentiment for all tweets. Even though the prediction accuracy is .9 this still implies that 10% of all tweets will be coded incorrectly. Therefore, results should be treated as indicative, also given the challenge of manually coding tweets. Results of applying Maxent to the total data set of tweets divided per day can be seen in Figure 28. The categories ‘Neutral’ and ‘Sadness’ are represented the most. ‘Anger’ increases from 4.4% to 6.0% and ‘Disgust’ from 0.9% to 1.3%. Tweets with the sentiment ‘Surprise’ had the smallest proportion with 0.5% and 0.7%.
Figure 28: Prediction of sentiment per day
22nd March 2016 23rd March 2016
3,5% 4,4% 0,7% 2,9% 0,5% 0,9% 1,3% 2,2% 2,4% 6,0%
40,7% 39,1% 49,3% 46,0%
Looking at the change over time on the first day it can be seen that the proportion of Tweets expressing sadness increased from 34% to 60% (see Figure 29). On the second day this trend is in the opposite direction from 53% to 35% (see Figure 30).
64/211 D2.4: Impact of social media, Version V1, PU Figure 29: Prediction of Sentiment per hour on 22nd March 2016
100% 11% 11% 9% 9% 90% 12% 14% 13% 12% 11% 11% 13% 13% 11% 80% 70% 34% 37% 41% 42% 47% 45% 49% 60% 63% 54% 54% 56% 66% 60% 50% 40% 30% 55% 51% 20% 45% 45% 40% 44% 40% 35% 33% 32% 10% 29% 25% 29% 0%
Neutral Sadness Other
Figure 30: Prediction of Maxent per hour on 23nd March 2017
100% 11% 90% 11%14%14%15%19%16%12%11%11%13%13% 11%11%13%13%13%14%14%12%14% 80% 30% 70% 38% 37% 35%34%34%35%37%34%34%40%35% 50% 40% 60% 53%50%48%45%44%45%51% 50% 40% 30% 59% 51%46%50% 53%54%53%51%50%52%52%48%51% 20% 36%36%38%40%37%39%37%38% 10% 0%
Neutral Sadness Other
We further created three word clouds based on tweets of three particular hours: 10am to 11am on the 22nd, 6pm to 7pm on the 22nd and 12pm to 1pm on 23rd March 2016. This throws further light on the development of the content of tweets over time. The word clouds reinforce the observation from above that feelings of sadness become more predominant throughout the first day. The first word cloud shows that apart from Brussels frequently used words are ‘news’, ‘anyone’, ‘makes’, ‘going’, ‘watching’, ‘all’ and ‘airportmetro’ (Figure 31). This are in general more expected to be part of tweets coded as ‘Neutral’ than as ‘Sadness’, however, words such as ‘love’ and ‘sad’ are also frequently used in tweets. Still, there is a clear difference compared with the second word cloud (Figure 32). The second word cloud shows that words such that emotive / relational words such as ‘Thoughts’, ‘families’, ‘people’, ‘Another’, ‘tragedy’ and ‘relative’ were used most frequently
65/211 D2.4: Impact of social media, Version V1, PU in addition to ‘Brussels’. These words are more associated with the expression of Sadness than reporting of news or information. The last word cloud shows that only ‘Brussels’ and ‘attacks’ are very frequently used - along with the emergence of the word ‘suspect’ associated with the police manhunt of those involved in the attacks (see Figure 33).
Figure 31: Word cloud of tweets on the 22nd between 10am and 11am
66/211 D2.4: Impact of social media, Version V1, PU Figure 32: Word cloud of tweets on the 22nd between 6pm and 7pm
Figure 33: Word cloud of tweets on the 23rd between 12pm and 1pm
67/211 D2.4: Impact of social media, Version V1, PU Social media roles In the early phases of the emergency, a complex interaction between citizens, traditional media outlets, and state agencies played an important role in amplifying messages. During the information phase of the emergency this involved different online actors re-posting messages from the Crisis Centre or embedding images of tweets and messages within the body of news bulletins and stories. Crucial in this process was the role of media organisations with large followings on sites such as Twitter and Facebook. This process led to a rapid dissemination of information as the wider networks of the media organisations allow information to spread beyond those who follow state actors. A key role for citizens was that of ‘citizen journalists’, providing ‘on-the-scene’ information about the events. The very first reports from the scene came from eye-witnesses on the ground and were disseminated in the form of pictures, videos and other messages [ASW17]. Given the dynamic nature of interactions online there is not a clear distinction between citizens, professional journalists and state actors, each disseminating information to their networks and re-posting information from other sources. As well as citizens, media and state actors posting information on social media the Crisis Centre attempted to use this data to inform their response. On the day of the attacks, a team of six was responsible for the online interactions on Twitter, Facebook and the website. In total this team sent 220 messages on Twitter and ended up with 50,000 new followers on Twitter and 8,000 on Facebook (organisational case study). Personnel within the centres drew on a range of media including press, broadcast and social media. Crisis Centre staff noted that social media is particularly important in gleaning information about events on the ground in the first hours due to its immediacy (ref – organisational case study). As time passed, the Crisis Centre continued to monitor social media in order to understand the immediate concerns of citizens in order to best respond to them. Operational decisions were also informed by social media data analysis with the decision taken to allow the transport system to restart after it was noticed from images posted online that large crowds were appearing outside stations that might form further targets for attacks. Citizens contributed in a number of ways including offering sympathy and solidarity with victims and in cases where citizens were local to the events offering practical help to those affected. In this mode some citizens can be characterised as ‘digital helpers’. Particularly the hashtag “PorteOuverte” (“open door”) was disseminated over Twitter to try to connect those in need of temporary accommodation with those able to provide it (see Figure 34). Tweets offering help numbered more than 3000 in the first 24 hours after the attacks with over 80 houses being opened [Alap16]. While this is a relatively modest proportion of the overall Twitter traffic given the scale of the event in these first hours reactions to these offers were very positive, showing that whether or not the offers were taken up they also acted as statements of solidarity and sympathy and were circulated widely.
68/211 D2.4: Impact of social media, Version V1, PU
Figure 34: Examples of “PorteOuverte”
Conclusion The Bombings in Brussels in March 2016 saw social media play a number of roles in the emergency response. In total, over 2 million tweets were posted about the attacks in the hours and days following the events. Social media was involved both in state communication with the public and used by emergency services to better understand the situation and informed their response. Citizens used social media during the incident in a range of ways both to express reaction to the events, solidarity with its victims and offers of help to those affected. In this last respect social media could be seen to form part of the wider social response to the emergency and to cater for the practical and emotional needs of the society. In keeping with both the 6-Stage model suggested by EENA paper and our own sentiment analysis, the early stage after an incident appears to be characterised by a large volume of informational posts with relatively little emotional content. As time passes the volume of such posts diminishes and is replaced by a high volume of tweets with emotional content expressing predominantly sadness and shock as well as support and solidarity to the victims.
69/211 D2.4: Impact of social media, Version V1, PU The Boston bombings This case study provides an example of how social media can interact with local emergency management practices and plans previously prepared in a man-made emergency. Additionally, as a terrorist attack, some learning could be found that complements the analysis of the Brussels attack.
Location and boundaries of the emergency Boston is an historic city which represents the economic and cultural heart of New England (Unites States). Its population of 650,28121 and 8 square miles (124 km2) form part of a larger metropolitan area known as Greater Boston, populated by 4.7 million people in 2014 and ranked as the tenth-largest such area in the US [WWW19]. Boston is located in the North part of the East Cost of the country, as Figure 35 shows. Figure 35: Location of Boston in the United States
Source: Google maps https://www.google.es/maps/place/Boston,+Massachusetts,+EE.+UU./ Among the cultural activities organised in the city, the Boston marathon is one of the most famous events. It is organised annually since 1897 [HKK14] and it is one of the six World Marathon Majors. About 30,000 runners are registered annually on average and the marathon attracts about half a million spectators [WWW20].
21 2011-2015 American Community Survey 5-Year Estimates (United States Census Bureau).
70/211 D2.4: Impact of social media, Version V1, PU Figure 36: Boston and municipalities around
Source: Google maps https://www.google.es/maps/place/Boston,+Massachusetts,+EE.+UU./
Moreover, Boston is considered one of the most prepared cities for emergencies, as it is one of the four U.S cities with an all-hazards plan [HKK14]. This plan includes the coordination among multiple government agencies when a man-made disaster, as the one analysed in this case study, happens. These are the main characteristics of the scenario where the bombings happened on April 15th 2013 and the manhunt of the expected culprits over the next days. The particular location of each of the main events over this emergency is pointed out in the map in Figure 36.
Main events and key incidents On April 15th 2013, two improvised explosive devices were set off close to the finish line of the 117th Boston Marathon. The first explosion happened at 2:49 pm, and the second took place 13 seconds later, about three hours after the winners completed the course. At that time, approximately 17,000 runners had completed the race while almost 9,000 were still advancing towards the finish line. The explosions occurred in the last 225 yards of the course (see number 1 in Figure 37), near a large number of spectators. The bombings caused three deaths and 264 people injured. The bombs exploded 200 yards apart, near the metal barriers that separated the spectators from the runners.22 Both bombs were made of
22 Federal Emergency Management Agency (2013).
71/211 D2.4: Impact of social media, Version V1, PU pressure cookers concealed in backpacks with low-grade explosives, nails, shards of metal, and ball bearings. President Barack Obama issued a disaster declaration two days following the bombings [HKK14].
Figure 37: Significant sites related to the Boston Marathon bombing of 2013.
Source: https://www.britannica.com/event/Boston-Marathon-bombing-of-2013 Encyclopaedia Britannica, Inc
It took some days for information on suspects and their goals to be clarified. However, the cause of the explosion was identified as a terrorist attack quite soon [BLRK15]. Over the next days, a manhunt of the two suspected terrorists took place. On April 18th at 5:10pm, three days after the explosions, the Federal Bureau of Investigation (FBI) released photographs and surveillance video of two suspects, enlisting the public’s help to identify them. This triggered a wave of speculation online, where members of the public were already working to identify the bombers from photos of the scene. Shortly after the photo release and a subsequent related shooting on the MIT campus” took place [SMOA+13]. Then, the patrol officer Sean A. Collier became the fourth victim of the attack when the Russian immigrants Tamerlan and Dzhokhar Tsarnaev tried to flee Boston [Morr16] (circle 3 in Figure 37) on April 18th 2013. The brothers hijacked a car at gunpoint (circle 4 in Figure 37) and the officials requested the residents of Boston and surrounding areas to remain indoors (i.e. a ‘shelter-in-place’ request). An urban lockdown was enforced in the Greater Boston Area, closing public transportation and most businesses on April 19th [Morr16]. A gun battle ensued between police officers and the two suspects. This resulted in one of the brothers, Tamerlan, to be shot and to be dragged along the road for several feet by a vehicle (circle 6 in Figure 37). Police officials found the remaining surviving suspect, Dzhokhar, hidden inside a resident’s boat in Watertown, Massachusetts (circle 7 in Figure
72/211 D2.4: Impact of social media, Version V1, PU 37) [HKK14]. The manhunt resulted in the death of one suspect and the capture of the other. The captured suspect Dzhokhar claimed that Tamerlan had been the mastermind of the attack moved by extremist Islamist beliefs, but that they acted independently of any formal terrorist organisation [HKK14]. He was charged of thirty counts relating to the Boston Marathon bombing and following incidents in Watertown, seventeen of which brought the possibility of a death sentence [Morr16]. The Boston Marathon bombings were considered the first successful terrorist attack on US soil since the September 11th, 2001 attacks [Garb15], increasing the feeling of vulnerability and the social impact of the bombings among the national and local community.
Patterns of social media usage After analysing available documentation it can be concluded that social media played an important role in the events which took place in Boston in 2013. Statistics show that internet users found social media as a key tool in this emergency, with millions of tweets in a few days, many of them using mobile phones, as well as other gadgets, both from the bombings scenes and from their homes and surroundings during the imposition of the lockdown order. Twitter and Facebook were the most analysed cases of how different types of social media were used to share information, manage the emergency and provide collective support in this emergency situation. The following sections analyse different purposes and patterns of use. Immediately following the Boston Marathon explosions, people close to the scene posted a flow of data on social media sites. This information was timely and individuals posted messages in Twitter which identified the location and a specification of the emergency events, even sooner than the public health authorities, which were alerted 9 minutes after the bombings [CCMC13].
73/211 D2.4: Impact of social media, Version V1, PU Figure 38: Cumulative time series of tweets from within a 35 mile radius of the Boston Marathon finish line selected using the stems “explode*”, “explos*” and bomb*” after the bombings at 2:49 p.m.
The coverage of the bombings and the ensuing four-day manhunt was widely covered through broadcasts live on television and via the internet [Morr16]. As there was a shelter- in-place in the city for most people within the Greater Boston Area, many citizens used social media to find out and share information about the events.
During the manhunt of the alleged culprits following the bombings, social media and the traditional media, especially television and radio, were used to be kept informed and updated [Morr16]. As there was a lockdown in the city, citizens relied on the different channels available to them in their homes, including social media. For example, “some residents tweeted photos of law-enforcement officers lying flat on neighbouring houses with machine guns drawn” [Morr16]. Thus, the situation of emergency and collective fear may have also been reinforced by the same channels that allowed citizens to be informed. While the Massachusetts Governor used traditional TV channels to advise citizens to be inside their homes [Morr16], citizens also used social media to amplify, update or share the information they considered relevant to their own and others’ security and awareness of the situation.
Sutton et al [SGSL+15] pointed out that Twitter is particular attractive for sharing messages that are not very complex and requiring only a few steps to understand the maximum amount of information (i.e. without links to other sites) during periods of imminent threat. Moreover, follower numbers are a strong predictor of retweets, so increasing connections is extremely valuable to share information, especially from official accounts whose
74/211 D2.4: Impact of social media, Version V1, PU information is considered reliable by the public and whose quality can be directly monitored by the emergency services.
The slogan "Boston Strong" appeared as a highly popular hashtag on Twitter two hours after the explosions in Cleveland (Ohio) far away from Boston. The hashtag “BostonStrong” had been used more than 1.5 million times one year after the bombings [Morr16] and it’s still used today. This slogan was present everywhere in Boston after the bombings, from t-shirts to cars, in tattoos or even mowed into the field on Fenway Park [Morr16]. The permanent use of this slogan has become a symbol of patriotism and survival, a mark of solidarity within a community which felt exposed facing this man-made emergency. Two college students used this sentence to stamp t-shirts for sale and donate 100% of the profits to the main marathon charity (One Fund) [Morr16] to benefit victims of the bombing. A Facebook group (@StayStrongBostonStrong) created for this aim on April 15th 2013 is still being used today, with more than 57,500 likes [WWW21]. This use of social media to provide economic support for the families affected by the events during Boston Marathon is another way of providing support. This campaign keeps working and it has now raised more than one million dollars [WWW22]. The official sources also used their Twitter accounts to transmit information and warning messages to the public about the Boston Marathon bombing and manhunt. Sutton et al. [SGSL+15] pointed out the relevance of the re-transmission of official messages to reach the widest audience and how the characteristics (content, style, and structural features) of those messages relate to a broader dissemination of the information, especially relevant under conditions of imminent threat. Although not all official agencies are emergency services, most of them are among official accounts. The use of social media, and Twitter especially, by the Local public officials, and the police in particular, registered relevant increases in attention as they provided real time updates from the scene of the bombing and additional messages through the following week.
Even though social media can be used to share false positive or misinformation, they can also be a really useful tool to raise timely alerts about specific events or emergencies as they happen [CCMC13]. This timely data, linked to geographical data, and in comparison to trends, can identify high spikes in keyword use, such as happens with “bomb*”, “explode*” or “explos*” in this case[CCMC13]. Moreover, in the long term, the impact of false positive can be reduced. So, these channels can complement traditional approaches to situation awareness after emergency events, such as in the Boston bombings.
One particular characteristic of the Boston bombings was that some emergency services were already located close to the bombing area assisting with the Marathon and so were ready to assist those in need after the explosions. Social media was therefore used both by the public to share timely updates and by emergency services to monitor such data and to provide advice and reassurance to the public [CCMC13]. For example, in the case of the American Red Cross, as they set up reception centres for those displaced in Watertown and in Cambridge, this organisation used their social media channels to engage with members of the general public and share information with them, as part of a “shelter in place” warning, to reduce citizens’ fears and concerns.
75/211 D2.4: Impact of social media, Version V1, PU Thus, given the particular characteristics of the Boston emergency, three factors have likely been impacting the use of social media and its interaction with emergency management: There were plans for all kind of hazards previous to the events that already considered the interaction among formal organisations at different levels. The bombing happened during a famous and very crowded marathon and emergency services were already and actively ready to act in case of emergency, even if the emergencies expected were others.
Social media roles The interaction between traditional media, especially television and radio, and social media created an “immense wave of public passion” during the bombings and the lockdown [Morr16]. This continuous sharing and dissemination of news, information and rumours relating to the events took place covering the different social media roles defined in D2.2. First, the community support and solidarity were perceived as a way to stay strong and increase resilience, potentially reducing the fear and trauma of these unexpected explosions, due to extremism ideas, in one of the most famous and traditional events in the city. One role adopted by social media posters, including citizens and official accounts, was ‘digital helper’ – providing both encouragement and reassurance to others. This included encouraging social cohesion and resilience-building, and reassuring that the government agencies were responding collaboratively. These messages were appreciated and retransmitted and helped to restore the sense of control, safety, predictability and trust that the terrorist event had challenged in many citizens and the Boston community. These tweets directed to the entire community helped to recover from the initial attack and shock while keeping alert to potential further terrorist activity [SGSL+15]. Moreover, social media were also used, in combination with direct engagement and use of traditional media, to raise funds for the bombing victims, and to organise volunteer support. Second, social media users played the role of ‘informants’: they acted as ‘citizen journalists’, sending photos and reporting the explosions and providing information during the manhunt. They shared what they saw with friends and family, emergency services and the local, national and international communities. Rumours, corrections and updates of the ongoing situation were posted and reposted through social media. Third, citizens also took on the role of ‘amplifier’ both in relation to the bombings and during the subsequent days of the manhunt. This had both positive and negative impacts. On the one hand, social media was used to spread and reinforce supportive messages in relation to the “Boston strong” campaign discussed previously. On the other hand, social media was also used to amplify negative messages towards Muslims and towards those with a physical resemblance of the two culprits. This means that during the trial process of the surviving terrorist there were concerns of bias as a result of such negative messages shared on social media and in the media [Morr16].
Quality of information One way of measuring the quality of the information shared by social media is considering the source of information. In this sense, social media messages originated by official sources are considered as quality ones and, thus, their higher dissemination and retransmission can help the spread of reliable information. Sutton et al [SGSL+15] pointed out that messages
76/211 D2.4: Impact of social media, Version V1, PU containing hazard impact, advisory and emotive/evaluative content are the strongest predictors of message retransmission, in addition to the number of followers and being directed tweets (i.e. those beginning with @name). However, some research pointed out that many social media users do not consider the origin of an idea or the credibility of the messenger and that many times opinions are taken as facts. For example, Reddit has an indicator of credibility of the contributors, but this has no impact on whether or not an individual’s post influenced the position taken by following contributors. In contrast, the more words they use or the more activity a comment produce and more individuals share similar information, push people to commit to that idea or message [GaKa16]. In the case of Reddit, the Boston Marathon bomber manhunt in Watertown was an example of how fast misinformation can spread. The authors examined five live thread discussions that took place before the identity of the bombers was made public, assessing 42,500 comments over a period of nine hours. These comments included emergence, dissemination and survival of false rumours. However, they were rejected once the true suspects were identified [GaKa16]. However, as noted in previous Deliverables (D2.2 and D2.3), there was evidence of the way social media has a self-correcting mechanism, as part of which false rumours are discredited by other social media users. Research by Starbird et al [SMOA+14] focused on misinformation (false rumours) shared through Twitter in the aftermath of the Boston Marathon bombing on April 15, 2013. Their analysis identified relevant themes and anomalies in the data. From the 100-most prevalent hashtags in the data, they created a network graph of relationships between them. (Each node represents a popular hashtag and is sized by the log number of times it appears in the set. Each edge connects two hashtags that appear in the same tweet and is sized by the log number of times they co-occur. They did not include the #boston hashtag in this graph, because it connected with every other tag.
Conclusions The Boston Marathon bombings in April 2013 and the following manhunt of the suspects implied a huge emergency in the city and the country, as it was recognised as the first terrorist attack on US soil since September11th, 2001. Emergency plans at local, state and federal level were implemented and social media played a significant role in the information and communication strategy from the official sources as well as among citizens. As such, it was seen as boosting the resilience-building of a whole community. The use of social media is widespread in general and also in emergencies. Sharing information and developing supportive actions, including volunteering management, were also very popular during the Boston emergency. Moreover, including social media channels in the formal strategies of emergency services and other official bodies have been more common in this man-made emergency than in the previous natural emergency cases studied on floods in the European Union. In this case, both official sources and citizens have participated and, although the strategic and anticipated expected use of social media by the emergency services was foreseen, the event increased the power and relevance of these instruments on the go.
77/211 D2.4: Impact of social media, Version V1, PU In addition to the wider use of social media over the years, the cause of the emergency, bombings in a famous marathon considered as a terrorist attack from the very beginning may also have had an impact on the scale of social media use. In this respect, the emotional needs of the wider community, and the use of social media to express the need for, and provide, support, meant that this case study is an example of a supportive social media intervention involving both citizens and official organisations. Regarding the quality of the information, this case study provided good evidence of the corrective facility of social media in correcting false rumours. In this sense, this case study presented the relevance of coordinated communication strategies in emergencies, across different channels and across different organisations working during the emergency. Although the geolocation of messages may have provided a very useful tool and a step forward in relation to using information from citizens by the emergency services, the review shows that there is still room for improvement in this regard.
78/211 D2.4: Impact of social media, Version V1, PU 7 Case studies of the EmerGent trials
Introduction This case study follows on from, and concludes, the last wave of case studies we implemented as part of the EmerGent impact assessment. These focused on the organisational dimension of emergency services’ use of social media. This last case study applies the framework we developed to two field trials of the EmerGent system at Dortmund and Hamburg Fire Services, carried out in the first half of 2017. This covered the second iteration of testing of the ESI interface, following an initial ‘scenario-based’ evaluation with representatives of emergency services, which led to the re-design of the ESI interface. The overall question we sought to answer was: “what is the current and potential impact of the EmerGent system?” (The methodology used for this case study is outlined in Section 2.5 above). This chapter presents the findings from this research by combining the results from both trials, as well as the simulation carried out in Salzburg in February 2017 (this explored the use of the ESI interface and app within the context of a simulated fire in a German city), to extract common themes and key learning as an important step towards the final section of this Deliverable which sets out the main conclusions and develops a contribution story for EmerGent.
Organisational backgrounds and use of social media Hamburg and Dortmund Fire Services are both large organisations. Hamburg fire service is the second largest fire service in Germany, employing 5,000 staff in various functions. It responds to around 280,000 incidents per year. Dortmund fire service consists of approximately 2000 professional and voluntary firefighters who respond to more than 115,000 incidents per year. Both organisations have Twitter and Facebook accounts which they use to send out alerts and instructions to the population in emergencies. Dortmund fire service also monitors social media in order to anticipate potential emergency situations and also to communicate with citizens, using the application Tweetdeck. For instance, during a recent Catholic Youth meeting they used social media to understand where groups are forming so they could send medical vehicles to the respective locations. In both organisations, social media use is somewhat centralised with use being formalised as part of communications roles. In addition, senior command staff in Dortmund use social media to permanently scan the situation, whilst in Hamburg staff can use social media if they have prior knowledge of use (there is no training programme). Neither organisation describes their use of social media as being ‘cutting edge’.
Motivations for trialling the EmerGent system Both fire services included only the EmerGent IT system in the trial. Testing the app was not possible in Dortmund because of strict security settings on work mobile phones. Motivations to take part differed slightly. Dortmund was keen to test whether the system was able to filter information in a way that improved their ability to anticipate emergencies and get a fast(er) impression of a situation on the ground (including receiving images). Hamburg fire service wanted to test whether the EmerGent system works and can improve
79/211 D2.4: Impact of social media, Version V1, PU their daily job as well as wanting to support the project and help improve the system so that it could be potentially helpful for fire stations.
Trial design The table below compares how the EmerGent trial was implemented in the two fire services. Both trials ran for a month and were implemented consecutively, starting with Dortmund then progressing to Hamburg (with Hamburg subsequently extended to also cover the G20 summit). In both cases, key words were defined before the trial started as required by the EmerGent system. The main differences in trial implementation were: The Hamburg trial involved a wider range of staff and use of key words Keyword definition was a discrete activity before the trial in Dortmund, whereas in Hamburg there was a process of ongoing refinement There were three major incidents in Dortmund during the trial period, whereas in Hamburg there were none.
80/211 D2.4: Impact of social media, Version V1, PU Table 6: Comparison of the Emergent Trials in Dortmund and Hamburg
Dortmund trial Hamburg trial
Duration 1 month 1 month
Scope EmerGent system EmerGent system
Participating Communications Communications departments Dispatch Dispatch Strategic planning Emergency staff on duty Field force
Key words used “fire”, “rescue” and “severe “fire/terror”, “CBRN”, weather” “malfunction of subway/bus”, “bad/contaminated food”
Key word definition Defined during 2-3 weeks Defined before the trial before the trial, using 1,000 tweets
Ongoing refinement of key words No ongoing refinement of key during the trial words
Supporting activities Creation of a user guide / Integration of EmerGent system manual of the EmerGent IT into dashboard also including system links such as radio, news, livestream cameras.
How integrated into EmerGent system used on the EmerGent used in addition to normal operations same machines and in parallel normal operations for periodic to other systems used. checks
Number of relevant Three major incidents No major incidents incidents during trial period
81/211 D2.4: Impact of social media, Version V1, PU Results of the trials The trials showed that the EmerGent system is stable and that the layout is clear and core functions self-explanatory. The fact that it is a web-based system has advantages for organisations operating under strict IT security protocols and hence may restrict very tightly what software can be downloaded onto work computers. The geo-tagging functionality makes it superior to other (free) solutions (e.g. Tweetdeck) and other than Google alerts, the system provides real-time information. Results from the Dortmund and Hamburg trials showed that the EmerGent system is useful for getting an overview of what’s going on. This is very helpful and relieves the pressure of looking at several platforms for relevant information at the same time. It is also helpful to observe reactions and feedback from citizens. Below we have listed some examples of the benefits the EmerGent system has shown during the trials. Box 1: Examples of usefulness of the EmerGent system from the Hamburg trial One employee in Hamburg Fire Service gave an example of a post that came in via EmerGent (though not before it came in via other ways of communication). The incident was a car accident with 5 to 6 vehicles involved and two injured people. As a result of the accident, streets were closed. The employee researched the links in the message and got more information by this means. This was useful in the sense of getting more and more detailed information. Information was not used in any particular way though in this instance, though one possibility could have been to navigate emergency vehicles around the closed streets. Also from the Hamburg trial came a few examples where the EmerGent system provided useful information. Examples were finding of a bomb, a fire and a closed tube. EmerGent messages on these topics were checked to see whether that information had already come in. Finally, a result from the Hamburg trial was also that those involved saw it as a good way to experiment more with social media (analysis) technology as the organisation as a whole is still not very advanced in this regard.
Learning This section sets out four key learning points from the two trials. The first learning point from the trial is that keywords need to be adjusted (continuously) to a particular context. The implementation of the two trials highlights the value of iteration in fine-tuning a prototype technology. The Dortmund trial showed that limiting the definition of key words to a set period before use of the system does not allow the system to fulfil its potential: incorrect configuration of the key words are the most likely explanation for the fact that the EmerGent IT system did not pick up much relevant information related to the three incidents that took place during the trial month. This learning was then taken into the subsequent Hamburg trial where there was an ongoing fine-tuning of key words leading to more (geographically) relevant information. So whilst at the beginning of the Hamburg trial a high volume of irrelevant messages were produced, constant refining of keywords added value by reducing this number.
82/211 D2.4: Impact of social media, Version V1, PU Second, what functionalities are needed and how such tools would be used depends on the target group. The communication team, for example, are not interested in specific messages, instead they want to get an overview and send an accurate representation of the situation to the public. The time factor is less important for the communication team whereas the dispatch centre needs information about an incident immediately and they are also interested in specific information. Third, both trials suggest that introducing the system into an emergency services organisation is likely to require some training. For instance, prior to the start of the trial Dortmund Fire Service produced a user manual for those likely to use the system. This was a user-friendly document telling staff how the system is used (e.g. what to click on). This was then sent to key staff who acted as multipliers of this information in the organisation. Those involved in the Hamburg trial suggested creating a video about how to use it and the functionalities and then offer a time where people can ask questions (e.g. via online chat). It is also important to be transparent about all functionalities, for example what does social media quality mean and how does it get assessed? Only when all functions are clear to the user is an effective usage possible. Organisation will still need continuous support. Finally, there is likely to be a need for local adaptation of the EmerGent system so that usage is in line with organisational processes and practices. Any guidance or training would therefore need to adapt generic tools such as a manual or video guidelines, e.g. by supplementing them with locally relevant information.
Future use Staff interviewed in both case study sites saw potential in the future use of the EmerGent system also because wider contextual factors (such as the current elevated level of violent extremist attacks and shortage of funding) make use of social media and free technologies to support analysis of information generated by them more interesting. Interviewees in both case study sites saw potential of the EmerGent system to connect citizens and emergency services more strongly as well as social media and the emergency management cycle because of the ability of the system to bundle messages from different platforms and hence offer an overview of a situation. It was noted that this partly depends on the role of the EmerGent user. Connecting citizens and emergency services is especially sought in public relations. It was further thought that EmerGent has potential to connect social media and other emergency service IT systems more strongly, also because it’s a web- based system which potentially offers value added (according to interviewees from the Hamburg field trial mainly to deliver improved situational awareness during an emergency). The post-project exploitation of the EmerGent IT system is likely to be accelerated if: There was an alarm function to alert users to incoming messages The system allowed the users to see which social media platform messages come from The duration with which messages are saved by the system was longer There was improved scope for the user to add key words and modify them (at the moment this process is centralised within EmerGent).
83/211 D2.4: Impact of social media, Version V1, PU Concluding summary The two field trials of the EmerGent system offered valuable opportunities to test the technology in practice. The experiences and results were somewhat different in the two sites, also because the Hamburg trial benefitted from important learning from the Dortmund experience about the importance of ongoing refinement of the key words. Nevertheless, there was agreement in both case study sites that the EmerGent system has potential for take-up and in relation to the overarching outcomes the project sought to achieve (further supported by the fact that the system has some advantages over other free solutions currently available). However, this exploitation was seen to be somewhat dependent on tweaking particular aspects of the system as listed above.
84/211 D2.4: Impact of social media, Version V1, PU 8 Conclusions Introduction: What is contribution analysis and what role does it play in EmerGent? A fundamental question for EmerGent – and in particular for work package 2 of the project – is “what is the impact of social media in emergencies?” As the project proposal and DoW observe ‘’the heart of the project EmerGent is the assessment of Impact of Social Media in Emergencies”. In this context, a key task of work package 2 was to develop a conceptual model, and tools, for assessing the impact of social media in general and, more specifically, of the three main products or outputs of the EmerGent project, which cover: A social media application (App) to help citizens to share information with ES more effectively before, during or after emergencies A system for the Novel Emergency Management in Social Media Generation for Information Gathering (IG), Information Mining (IM), establishing Information Quality (IQ), and Information Routing (IR) – eventually developed as the EmerGent ‘Emergency Services Interface’ (ESI) Guidelines on social media use in emergencies by ES and citizens. This work on impacts assessment therefore had three main purposes: A formative purpose - understanding what does and does not work when using social media in emergencies provided an opportunity to contribute to continuously improving EmerGent outputs A summative purpose – applying these understandings to what EmerGent has achieved at the end of the project enabled us to make an evidence-based judgement of ‘what works, for whom, under which circumstances’ A learning purpose – the research and evaluation results produced over the lifetime of EmerGent provided valuable lessons – in the short term, feeding into production of the EmerGent ‘Guidelines’ – and in the longer term contributing to more effective design and implementation of similar initiatives in the future, and in different application areas. It should be borne in mind that work package 2 was structured and shaped in the project DoW as a ‘research’ rather than an evaluation Task. ‘Evaluation’ was quite narrowly-defined in EmerGent as a ‘technical’ task carried out in work package 6, and aimed at reviewing the functionalities of the EmerGent ESI and App through iterative paper-based, live system demonstrations and field tests, in order to improve the usability, user-friendliness and efficiency and effectiveness of the ESI and App. This means there are restrictions on the inferences we can make about the impact of EmerGent. Much of the research carried out in EmerGent on the impact of social media has involved ‘indirect’ or ‘potential’ users of the EmerGent outputs – for example case studies of the use of social media in real emergencies; surveys of emergency services; surveys of citizens. Assessment of the EmerGent outputs themselves has been restricted to the ‘technical evaluation’ described above, plus additional evaluations of trials of the ESI and App in Hamburg and Dortmund, carried out as ‘case studies’ by the Tavistock Institute. It should also be noted that none of these tests covered the EmerGent ‘Guidelines’. Therefore, these assessments of the EmerGent outputs themselves can be described as a specific
85/211 D2.4: Impact of social media, Version V1, PU composite case study – based on the EmerGent ESI and App - of the use of EmerGent by emergency services, rather than an evaluation of EmerGent itself. This poses a big challenge for EmerGent impacts assessment. Many discussions of impact evaluation argue that it is essential to include a ‘counterfactual’ [LoRo12]. Counterfactual impact evaluation involves comparing the outcomes of interest of those who have benefitted from an intervention (the ‘treatment group’) with those of a group similar in all respects to the treatment group (the ‘comparison/control group’), but who have not been exposed to the intervention. The comparison group provides information on what would have happened to the participants in the intervention had they not been exposed to it. The classic – ‘gold standard’ – method for carrying out such a counterfactual impact evaluation is an ‘experimental’ approach based on using randomised controlled trials (RCTs) where participants are randomly assigned to the treatment and comparison groups and then common outcomes are compared before and after the intervention for both groups. In the EmerGent case, this would imply, for example, comparing an emergency service that had used the EmerGent ESI and App with a similar service that had not, on the basis of indicators like number of incidents detected; speed of response, and so on. However, the EmerGent project has not been designed with experimental impacts assessment in mind. We have not had the opportunity within the project to maintain the conditions required in RCTs, particularly ‘temporal priority’ - the assumption that a suspected cause precedes an event (for example, that the use of the EmerGent ESI has ‘caused’ an improvement in emergency services management of a crisis). This is because of factors often cited in the literature that inhibit the use of experimental methods - infrequent data sampling, non-existent baselines, a project with multiple interventions and complex spill-over effects; insufficient units of assignment to conduct tests of statistical significance [SSMF+12] [BBS14]. How can we therefore identify outcomes we can demonstrate are ‘caused’ by and can be attributed to EmerGent? Contribution analysis provides a solution to this challenge because it provides a way of assessing causal questions and inferring causality in interventions. It aims to create a causal chain – or ‘contribution story’ that links actions and events to outcomes. It explores attribution ‘through assessing the contribution a programme is making to observed results. It sets out to verify the theory of change behind a programme and, at the same time, takes into consideration other influencing factors’ [Toul10] [Mayn12]. In a nutshell, by developing a Theory of Change showing the links between the activities, outcomes and contexts of the project and collecting evidence from various sources to test this theory, the aim of Contribution Analysis is to build a credible ‘performance’ story of that project. The methodology used to carry out the contribution analysis for the EmerGent project is described in Section 2.6, while the following section sets out how it was implemented with regard to the project and its overall aims, objectives and Theory of Change.
Implementation of the Contribution Analysis Step 1: Circumscribing the issue – setting out the attribution problem to be addressed. This Step entailed specifying the ‘presenting problem’ EmerGent hoped to address; the underlying causes of the problem; the expected improvement or change EmerGent hoped
86/211 D2.4: Impact of social media, Version V1, PU to make to the problem, and the key evaluation – or ‘impact assessment’ – questions to be addressed. The presenting problem addressed by EmerGent, and the starting point of the project, is that social media’s potential contribution to managing emergencies, and hence its contribution to reducing damage and loss of life, has not been adequately harnessed. This problem is based on the theory (hypothesis) that, at present, emergency services and citizens are weakly connected and at the same time social media and emergency services are weakly connected. EmerGent seeks to promote a solution to this problem by developing systems, tools and services that will change the nature of the relationships between citizens, emergency services and social media. The overarching evaluation/impacts assessment question to be addressed is: What results has the EmerGent project achieved during its lifetime, and why? What can we conclude from these results about EmerGent’s likely medium and long term effects? For the purposes of the Contribution Analysis, this can be phrased as three key evaluation/impacts assessment questions: Did the use of social media in general strengthen the connection between emergency services and citizens and does this enhance safety, speed up recovery and ultimately save lives? Did EmerGent strengthen the connection between social media and the emergency management cycle (EMC) and did this improve the situational awareness of emergency services during emergencies? Did EmerGent strengthen the integration between social apps, the EmerGent IT system and emergency services IT systems and did this improve the situational awareness of emergency services during emergencies? Step 2: Developing the EmerGent Theory of Change The initial (baseline) Theory of Change for EmerGent was developed using the following methodology: A review of project documentation (including the proposal and the DoW) A literature review carried out in work package 2 of what happens when social media are used in emergencies, and how these effects and causes could be captured and measured – published in D.2.1 Review of the literature reviews on the use of social media in recent emergencies, carried out by project partners and published in D3.1 and D3.4 An online survey to explore the EmerGent consortium partners views on the likely outputs and outcomes of the project. This produced three representations of the EmerGent Theory of Change, at three different levels: A very high level logic model outlining in very broad terms how EmerGent is assumed to make a difference
87/211 D2.4: Impact of social media, Version V1, PU A more detailed logic model which draws on information from work packages and tasks in the DOW, offering a more granular picture about EmerGent activities and outputs and the outcomes and impacts they are likely to produce A very detailed breakdown of anticipated outcomes and impacts from EmerGent, which organises the Theory of Change into a logical sequence of short-term outcomes, medium term outcomes, longer term outcomes (impacts) and ultimate impacts. These three representations are set out in full in EmerGent Deliverable 2.1. From these three representations, a summary narrative of the Emergent Theory of Change can be constructed: The context shows there are weak connections between the use of social media by emergency services and citizens, between social media and the EMC and between social media apps and ES IT systems To address these weaknesses the activities (inputs) carried out by EmerGent include: research on the use and impact of social media in emergencies; research on the potential for citizen involvement and use of social media by emergency services; technical analysis of processes, tools and interfaces, and design of the system architecture These activities lead to the production of outputs that will offer new methods and tools – an IT system for novel Emergency Management and an App for citizens – as well as Guidelines to support the adoption of social media in emergency services The utilization of these outputs lead to outcomes that strengthen the connections between citizens and emergency services, social media tools and the Emergency Management Cycle (EMC) and between Social Apps, the EmerGent IT system and ES IT systems and hence lead to improved awareness of the situation during emergencies Ultimately, these outcomes will lead to longer term impacts of increased objective and perceived safety of citizens before, during and after emergencies, through improved prevention, response to and recovery actions by emergency services. For the purposes of the Contribution Analysis, the three baseline Theory of Change representations were streamlined to produce a ‘Summary Theory of Change’, provided in the Table and Figure below. These show the ‘external influences’ that have potential to affect the Theory of Change; the ‘assumptions’ – the events and conditions necessary for the causal links in the Theory of Change to occur - and the ‘risks’ that may prevent these assumptions occurring. They also include the ‘primary explanatory mechanisms’ underlying the Theory of Change. These ‘primary explanatory mechanisms’ are defined as ‘underlying entities, processes, or structures which operate in particular contexts to generate outcomes of interest’ [AsLe10]. They make a bridge between the project activities/outputs and its generated outcomes. So, for example, in EmerGent, we have research and development activities to produce the EmerGent IT system (activity/output); the ensuing mechanisms (emergency service software systems have integrated options for filtering and assessing information) and the outcome (Social apps, EmerGent IT system and ES IT systems are better or more strongly integrated).
88/211 D2.4: Impact of social media, Version V1, PU Table 7: EmerGent Summary Theory of Change External Technological change Influences Policy Environment Commercial market trends Leaders & Influencers in ES Social Media cultural trends
Activities A1: Emergent Research, Development & Design Activities
Outputs O1: EmerGent Guidelines O2: Emergent IT system O3: EmerGent Social App
Mechanisms M1: ‘Awareness-Reach-Trust’. Citizens and ES increased awareness of each other’s activities, behaviours and information needs improves the quality of information shared;
increases ES use of social media before, during and after an emergency; reduces dependence on traditional media and reaches a larger number of people; reduces negative impact of misinformation and establishes trust in social media M2: ‘Process-Effectiveness-Embedding’. ES integrate social media into their workflows or create new processes to do so, which allows them to respond more effectively to incidents and leads to better understanding by citizens of the benefits of using social media in emergencies, so that their use becomes standard practice, and citizens share information and act as first responders. M3: ‘Workflow-Functionality-Awareness’. By integrating social media into the ES workflow, social media data are used more often during emergencies and more information is posted by ES; functionality for assessing a situation is better integrated and situational awareness improves.
Assumptions A1. Emergency services will reach audiences more efficiently using social media A2. Citizens are homogenous in their use of social media A3. ES are homogenous in their use of social media A4. People will use an EmerGent App A5. Social Media are homogenous A6. Social Media services are stable
Risks R1. Social media reduces effectiveness and efficiency of ES R2. Citizens use social media unpredictably R3. ES use of social media is variable R4. People are not interested in/can’t access the App R5. Different people use different social media for different purposes R6. Technical change outpaces EmerGent
Intermediate OC1: Citizens and ES more strongly connected Outcomes OC2: Social media tools are better connected to ES OC3:Social App, EmerGent IT and ES IT are better integrated
89/211 D2.4: Impact of social media, Version V1, PU
Ultimate I1: Enhanced Citizen Security & Safety goals/ I2: Improved Situational Awareness Impacts
90/211 D2.4: Impact of social media, Figure 39Version: Emergent V1, PU Summary Theory of Change
91/211 D2.4: Impact of social media, Version V1, PU Step 3: Gathering existing evidence on the Theory of Change The evidence to interrogate the Theory of Change was collected and analysed from the following sources: A survey of citizens’ attitudes to and use of social media, with a focus on use of social media in emergencies A survey of emergency services’ attitudes to and use of social media A case study of the use of social media in the ‘London Riots’ of 2011 Case studies of the use of social media in flooding in different locations and timeframes, i.e.: Germany (2013); Georgia (2015); Slovenia (2014); Poland (2010); Western Norway (2014); UK (2013-14). The following Tables show the evidence collected and analysed for each of the primary explanatory mechanisms specified in the Theory of Change, covering i) the evidence that supports the explanatory mechanism ii) the evidence that refutes the explanatory mechanism, together with estimations of the strength of this evidence, as well as evidence on the ‘assumptions’ embedded in the Theory of Change.
Table 8: Supporting and Refuting Evidence
Mechanism 1: ‘Awareness-Reach-Trust’: Citizens and ES increased awareness of each other’s activities, behaviours and information needs improves the quality of information shared; increases ES use of social media before, during and after an emergency; reduces dependence on traditional media and reaches a larger number of people; reduces negative impact of misinformation and establishes trust in social media
Supporting Evidence Strength
London riots supported co-operation between citizens and ES through discouraging some Rather strong people from participating, organizing citizens in cleaning up the damage, raising money to
compensate those negatively affected
Flooding case studies show evidence of social media contributing to ES-citizen co-operation on informing, alerts, response co-ordination. Rather strong
Flooding case studies show that social media are efficient and effective in broadcasting messages to particular groups of citizens and the media, even more so if a sufficient number Rather strong of followers has been accumulated before the emergency and these followers contain citizen
amplifiers
Refuting Evidence Strength
The flooding case studies show that many emergency services did not have the structures, Rather strong processes, skills and resources to use information from social media effectively during the floods.
92/211 D2.4: Impact of social media, Version V1, PU
Mechanism 2: ‘Process-Effectiveness-Embedding’. ES integrate social media into their workflows or create new processes to do so, which allows them to respond more effectively to incidents and leads to better understanding by citizens of the benefits of using social media in emergencies, so that their use becomes standard practice, and citizens share information and act as first responders.
Supporting Evidence Strength
After the London riots, the Met Police put in place a number of organisational and practical Rather weak changes in how they used social media in daily operations
The flooding case studies show in several of the cases that ES learning about needs and Rather strong familiarity with relevant technologies prompted a shift in how they view and / or use social media during and after each emergency, so that using social media to broadcast information has become part of the emergency management routine.
Refuting Evidence Strength
The flooding case studies show that many emergency services did not have the structures, Rather strong processes, skills and resources to use information from social media effectively during the floods
Mechanism 3: ‘Workflow-Functionality-Awareness’. By integrating social media into the ES workflow, social media data are used more often during emergencies and more information is posted by ES; functionality for assessing a situation is better integrated and situational awareness improves.
Supporting Evidence Strength
None of the emergency services in the flooding case studies have had technologies in place None to fully harness the information generated in social media during emergencies.
Refuting Evidence Strength
Flooding case studies showed instances where social media information overload Moderately contributed to reduced functionality and integration of ES IT systems strong
The London riots case shows that turning information into intelligence is more challenging Moderately during extreme events resulting in millions of tweets and other social media data in a short strong period of time
93/211 D2.4: Impact of social media, Version V1, PU Table 9: Assumptions Assumption Supporting/Refuting Evidence
A1. Emergency services will reach Flooding case studies show that efficient reach comes with the audiences more efficiently using social trade-off of receiving greater citizen responses which can place media an additional burden on emergency services – the way they respond then affects the mood of citizens
A2. Citizens are homogenous in their use of London riots show citizens take up different roles during social media emergencies, including informant, helper, amplifier and citizen journalists. Particular types of social media are more likely to be used by certain demographic and socio-economic profiles for particular purposes. Flooding case studies show that citizens take up different roles in social media during emergencies, and they also use different technologies
A3. ES are homogenous in their use of Flooding case studies show that emergency services differ in social media their social media skills and knowledge as well as their organisational context. Organizations vary in the speed and enthusiasm with which they adopt social media. Some are resistant to innovation
A4. People will use an EmerGent App Flooding cases show that current emergency apps even when focused on specific localities tend to have relatively low download and use rates.
A5. Social Media are homogenous Flooding case studies show that social media are used for different purposes by different people, and have cultural specificities (‘cultural protocols’).
A6. Social Media services are stable Flooding case studies show that social media are fragile (they come and go).
The next set of Tables present the ‘Alternative primary explanatory mechanisms’ together with the ‘Influencing Factors’ identified using the Relevant Explanation Finder (REF) methodology. Each alternative explanation and influencing factor is described using six dimensions: Description of the mechanism Type of explanation or factor identified. This covers four types: o Primary explanation (mechanism) – a mechanism identified and purported to be the target intervention mechanism that accounts for and explains the observed outcomes (i.e. the primary explanatory mechanism in the contribution story). o Direct rival (mechanism) – a mechanism, different from the target intervention mechanism, that accounts for and explains the observed outcomes (i.e. undermines the contribution story). o Commingled rival (mechanism) – other mechanisms, along with the target
94/211 D2.4: Impact of social media, Version V1, PU mechanism, that both contribute to and explain the observed outcomes (i.e. revise the contribution story). o Implementation rival (factors) – influencing factors in the implementation process, not substantive intervention mechanisms, that modify the outcomes (i.e. revise the contribution story). Explanation level. This also covers four types: o The individual level – the capacities of the key actors and stakeholders such as interests, attitudes, capabilities and the credibility of (for instance) professionals or beneficiaries. o The interpersonal level – the relationships required to support the intervention, such as lines of communication, management and administrative support, union agreements and professional contracts. o The institutional level – the setting in which the intervention is implemented, such as the culture, leadership etc. of the implementing body. o The wider (infra-) structural level - political support, the availability of funding resources etc. Identifiers – provides descriptions of the possible identifiers that offer proof or disproof of the existence of the various influencing factors and alternative explanations. Degree of influence – this summarizes the degree of influence of both the factors identified and the underlying mechanisms for the observed outcomes. It covers five attributes to be considered in assessing the degree of influence: o Certainty: The degree to which the observed outcome pattern matches the one predicted by the factor or mechanism. o Robustness: The degree to which the factor or mechanism is identified as a significant contributor across a broad range of data sources and data collection methods. o Range: The degree to which the factor or mechanism contributes to a broad range of the outcomes of interest. o Prevalence: The degree to which the factor or mechanism contributes to the outcomes of interest across a wide range of implementation environments and target groups (e.g. different implementation sites and / or types of intervention). o Theoretical grounding: The factor or mechanism is informed by theory (identifies existing theories of which it is an example) and is cast in specific terms (i.e. it is not vague). Implications – this provides conclusions regarding the implications of the factors and mechanisms for the contribution story.
95/211 D2.4: Impact of social media, Version V1, PU Table 10: Alternative primary explanatory mechanisms M1A
Alternative M1A: Citizens and ES less strongly connected through social media Explanation
Mechanism Citizens and ES increased awareness of each other’s’ activities leads to fragmentation of behaviours, sometimes in oppositional ways to those intended
Type Direct Rival
Level Individual
Identifiers London Riots case study shows social media supported the spread of the riots themselves and helped coordinate activities of rioters.
Flooding case studies show social media plays complementary or subordinate role to conventional media. Flooding case studies show different levels of co-operation between citizens and ES, with some ES adopting ‘top-down’ approach; some citizens not sharing with ES. Flooding cases show although many citizens shared information, many ES did not use it because of access issues, information overload or validity issues.
Degree of Influence
Certainty Moderate. This patterns broadly fits the alternative explanation
Robustness Low. Low triangulation with other data sources
Range Low. The behaviours cited are only two of a large range of contributory factors
Prevalence Low. The explanation relates to sub-groups of the target population
Grounding Low. There is very little evidence in the literature to support this explanation
Implication The evidence that the use of social media weakens the connection between citizens and emergency services is weak. However, the evidence does suggest that neither citizens nor emergency services are homogenous groups. Citizens take up different roles during emergencies; including informant, helper, amplifier and citizen journalists - and they also use different technologies. Particular types of social media are more likely to be used by certain demographic and socio-economic profiles for particular purposes. Emergency services differ in their social media skills and knowledge as well as their organisational context. Different types of organisation adopt different bridging and engagement strategies with citizens.
96/211 D2.4: Impact of social media, Version V1, PU Table 11: Alternative primary explanatory mechanisms M2A
Alternative M2A: Parallel lines Explanation
Mechanism Social media runs as a parallel activity within ES systems, and is utilised by individuals or small groups as a ‘bonus feature’ without being integrated, thereby restricting the capacity of ES to respond more effectively to incidents and leading to citizens not fully understanding the benefits of using social media in emergencies
Type Direct Rival
Level Institutional-interpersonal
Identifiers The flooding case studies show that many emergency services did not have the skills and resources to use information from social media effectively during the floods.
Flooding case studies show lack of organisational structures and procedures in place to support the use of social media
Degree of Influence
Certainty Moderate. This patterns broadly fits the alternative explanation
Robustness Low. Low triangulation with other data sources
Range Low. The pattern shown is only one of a large range of contributory factors
Prevalence High. Case studies show this pattern was repeated across different ES
Grounding Low. There is very little evidence in the literature to support this explanation
Implication The introduction of social media into ES organisational systems and practices can be disruptive. To strengthen connections between social media tools and ES systems requires relevant skills and new organisational systems and processes. There is some evidence that some emergency services will be resistant to these disruptive changes, at some levels, leaving social media as an ‘outsider’ or ‘exotic’ feature that is utilised only by social media ‘champion’s and innovators.
97/211 D2.4: Impact of social media, Version V1, PU Table 12: Alternative primary explanatory mechanisms M3A
Alternative M3A: Social media analysis tools contribute to system dysfunctionality Explanation
Mechanism Added flow of information overloads ES IT systems and hampers functionality and situational awareness is decreased
Type Direct rival
Level Institutional
Identifiers London riots case study shows turning information into intelligence is more challenging during extreme events resulting in millions of tweets and other social media data in a short period of time
Flooding case studies showed instances where social media information overload contributed to reduced functionality and integration of ES IT systems
Degree of Influence
Certainty Moderate. This patterns broadly fits the alternative explanation
Robustness Moderate. Triangulation of data sources show recurring pattern
Range Low. The pattern shown is only one of a large range of contributory factors
Prevalence High. Case studies show this pattern was repeated across different ES
Grounding Low. There is very little evidence in the literature to support this explanation
Implication There is some evidence to cast doubt on EmerGent’s third expected (desired) outcome – that the Social App, EmerGent IT and ES IT are better integrated because, by integrating social media into the ES workflow, social media data are used more often during emergencies and more information is posted by ES; functionality for assessing a situation is better integrated and situational awareness improves. What seems to be a repeated pattern is that integrating information from social media apps and other sources within ES control systems can disrupt ‘normal’ information flow and lead to system overload. This could lead to the opposite result to that intended – a situation in which situational awareness in ES control rooms is undermined.
98/211 D2.4: Impact of social media, Version V1, PU Table 13: Influencing Factor - Cultural Contextualisation
Influencing Factor Cultural contextualisation
Mechanism Familiarity with and routine embedding of social media in individual daily life will increase the probability that citizens will use social media to engage with other citizens and ES in emergencies and will shape their use behaviours
Type Implementation rival
Level Interpersonal
Identifiers Flooding case studies show that social media are used for different purposes by different people, and have cultural specificities (‘cultural protocols’). Moreover, the trajectory of their development is shaped by use (‘value embedded action’). It also matters whether social media are open or closed, and how people view them depends very much on who they are choosing to follow
Flooding case studies show that a key variable determining use of social media in emergencies is already using it in normal daily life.
Degree of Influence
Certainty Moderate. The case study evidence, supported by the literature review and a 2014 survey on citizens use of social media, shows that regular users of social media are more likely to use it in emergencies.
Robustness Moderate. The mechanism was identified as contributing factor in literature reviews, the Citizens Survey and the case studies.
Range Low. The explanation only accounts for one possible behavioural factor in citizens use of social media in emergencies
Prevalence Low. Additional occurrences of this factor are required in a wider range of environments.
Grounding Moderate. There is some evidence from ‘value embedded action theory’ (Cullen and Cohen, 2007) that technologies have cultural specificities and that the trajectory of their development is shaped by use (‘value embedded action’).
Implication The evidence suggests that EmerGent is most likely to achieve its intended outcome of developing a stronger connection between citizens and emergency services by improving citizens’ awareness of emergency services information needs (both in terms of content and quality) among those already using social media. It also implies that that the strategies ES devise to maximise the potential of social media need to reflect a typology of different cultural protocols. Conversely, ES cannot assume that all citizens uniformly will respond in
99/211 D2.4: Impact of social media, Version V1, PU the same way to the same messages. Another implication is that if the EmerGent guidelines are going to effect change, they have to cater for a range of experiences and needs. Finally, in order to help emergency services use the tools provided by EmerGent they need to be provided with help and guidance on how to do this.
100/211 D2.4: Impact of social media, Version V1, PU Table 14: Influencing Factor - Organisational Culture
Influencing Factor Organisational culture
Mechanism The impact of SM information on ES management of emergencies depends on key organisational culture factors like the efficiency of ES info-management systems; skills base and experience of using social media
Type Implementation rival
Level Institutional
Identifiers Flooding case studies show that efficient reach for ES with social media comes with the trade-off of receiving greater citizen responses which can place an additional burden on emergency services – the way they respond then affects the mood of citizens.
The London Riots and the Flooding case studies show that ES often lack the practice and skills of identifying and interpreting relevant information.
Degree of Influence
Certainty Moderate. The case study evidence shows this pattern is found across a range of emergency services.
Robustness Moderate. The mechanism was identified as contributing factor in literature reviews, an ES Survey and the case studies.
Range Low. The explanation only accounts for one possible behavioural factor in ES use of social media in emergencies
Prevalence Low. Additional occurrences of this factor are required in a wider range of environments.
Grounding Moderate. There is an established body of evidence form organisational theory that supports the concept of organisational resistance.
Implication The evidence suggests that the use of social media is not yet common practice for all emergency services and they often lack the practice and skills of identifying and interpreting relevant information. This implies that the scale of EmerGent impact in relation to emergency services broadcasting use of social media is likely to vary depending on existing organisational practices. One of the key challenges emergency services have at present is accessing and making sense for response purposes of the increased volume of social media information that is generated during emergencies. In none of our case studies did emergency services have a way to systematically access social media data. It would therefore seem that here lies a great potential impact for EmerGent.
101/211 D2.4: Impact of social media, Version V1, PU Step 4: Assemble the Contribution Story On the basis of the analysis carried out in Step 3, the initial Contribution Story for EmerGent is presented below. It provides a narrative to explain how and why EmerGent’s expected outcomes are likely to be caused by the sequence of events and actions carried out in the project. It also specifies the weaknesses in the story, in light of the available evidence, and the relevance of the other influencing factors. It should be acknowledged that this preliminary contribution story is not based on the measurable outcomes achieved by the project. As noted above, because the opportunities for evaluating the direct outputs – and their associated outcomes – of EmerGent have been restricted by the way the project and its DoW has been constructed, our Contribution Analysis explores the broader plausibility and ‘performance’ of the underlying ‘Theory of Change’ that underpins the project. It assesses the plausibility of the ‘primary explanatory mechanisms’ that underpin EmerGent’s Theory of Change and which are assumed to lead to the three desired and expected outcomes of the project, i.e. Citizens and ES are more strongly connected; Social media tools are better connected to the EMC; Social apps, EmerGent IT system and ES IT systems are better or more strongly integrated. Strengths of the Contribution Story EmerGent’s first expected outcome – that Citizens and ES are more strongly connected through social media in emergency situations – appears to be supported by the evidence. This outcome requires evidence to support the following explanatory mechanism: that citizens and ES increased awareness of each other’s activities, behaviours and information needs improves the quality of information shared; increases ES use of social media before, during and after an emergency; reduces dependence on traditional media and reaches a larger number of people; reduces negative impact of misinformation and establishes trust in social media. The London Riots and flooding case studies support this explanatory mechanism. They provide strong evidence that demonstrate co-operation between citizens and ES in terms of providing information and reassurance; disseminating alerts; organizing groups of citizens to support the ES response. The evidence shows that social media are efficient and effective in broadcasting messages to particular groups of citizens and the media, even more so if a sufficient number of followers have been accumulated before the emergency and these followers contain citizen amplifiers. EmerGent’s second expected outcome – that social media tools are better connected to the EMC – requires evidence to support an assumed mechanism in which ES integrate social media into their workflows or create new processes to do so, which allows them to respond more effectively to incidents and leads to better understanding by citizens of the benefits of using social media in emergencies, so that their use becomes standard practice, and citizens share information and act as first responders. The evidence to support this explanation is mixed. The London riots case study showed that, after the riots, the Met Police put in place a number of organisational and practical changes in how they used social media in daily operations. Similarly, the flooding case studies show in several of the cases that ES learning about needs and familiarity with relevant technologies prompted a shift in how they view and / or use social media during and after each emergency, so that using social media to broadcast information has become part of the emergency management routine. However, there is no clear evidence to show that emergency services in general have the mechanisms and systems in place to integrate social media into their workflows.
102/211 D2.4: Impact of social media, Version V1, PU EmerGent’s third expected outcome – that Social apps, the EmerGent IT system and ES IT systems are better or more strongly integrated – requires evidence to support an assumed mechanism in which, by integrating social media into the ES workflow, social media data are used more often during emergencies and more information is posted by ES; functionality for assessing a situation is better integrated and situational awareness improves. There is no evidence to support this explanation. Weaknesses of the Contribution Story The counter-evidence against EmerGent’s first expected outcome – that Citizens and ES are more strongly connected through social media in emergency situations – is relatively strong, but limited, and is based on evidence from the flooding case studies that shows that many emergency services did not have the structures, processes, skills and resources to use information from social media effectively during the floods. The evidence also undermines some of the assumptions underlying this explanation. The assumption that ‘emergency services will reach audiences more efficiently using social media’ is set against evidence from the flooding case studies which show that efficient reach comes with the trade-off of receiving greater citizen responses which can place an additional burden on emergency services – the way they respond then affects the mood of citizens. The assumption that ‘citizens are homogenous in their use of social media’ is set against evidence to show that citizens take up different roles during emergencies, and that particular types of social media are more likely to be used by certain demographic and socio-economic profiles for particular purposes. The evidence shows that social media are used for different purposes by different people, and have cultural specificities. The counter-evidence against EmerGent’s second expected outcome - that social media tools are better connected to the EMC – is relatively strong. The flooding case studies show that many emergency services did not have the structures, processes, skills and resources to use information from social media effectively during the floods. The evidence also undermines some of the assumptions underlying this explanation. The assumption that ‘ES are homogenous in their use of social media’ is not supported by the evidence. The evidence shows that emergency services differ in their social media skills and knowledge as well as their organisational context. Organizations vary in the speed and enthusiasm with which they adopt social media. Some are resistant to innovation With regard to EmerGent’s third expected outcome – that Social apps, the EmerGent IT system and ES IT systems are better or more strongly integrated – the evidence shows that none of the emergency services in the flooding case studies have had technologies in place to fully harness the information generated in social media during emergencies. In contrast, the flooding case studies showed instances where social media information overload contributed to reduced functionality and integration of ES IT systems. The London riots case showed that turning information into intelligence is more challenging during extreme events resulting in millions of tweets and other social media data in a short period of time. The evidence also undermines some of the assumptions underlying this explanation. The assumption that ‘People will use an EmerGent App’ is not borne out by the evidence. The assumption that ‘Social Media services are stable’ is also not supported. The evidence shows that social media are fragile and their attributes and use patterns change rapidly.
103/211 D2.4: Impact of social media, Version V1, PU Alternative explanations and Influencing factors The contribution analysis identified three possible rival explanatory mechanisms to the three assumed mechanisms embedded in EmerGent’s Theory of Change. It should be noted that – because the EmerGent contribution analysis works with ‘expected’ rather than ‘actual’ outcomes – these rival explanations are also based on ‘possible’ expected outcomes, rather than actual observed outcomes. The three possible outcomes identified, and their associated explanatory mechanisms are: Citizens and ES less strongly connected through social media - citizens and ES increased awareness of each other’s’ activities leads to fragmentation of behaviours, sometimes in oppositional ways to those intended. Parallel lines - social media runs as a parallel activity within ES systems, and is utilised by individuals or small groups as a ‘bonus feature’ without being integrated, thereby restricting the capacity of ES to respond more effectively to incidents and leading to citizens not fully understanding the benefits of using social media in emergencies. Social media analysis tools contribute to system dysfunctionality – the added flow of information overloads ES IT systems and hampers functionality and situational awareness is subsequently decreased. In the case of the first rival explanation, the evidence that the use of social media weakens the connection between citizens and emergency services is weak. However, the evidence does suggest that neither citizens nor emergency services are homogenous groups. Citizens take up different roles during emergencies; particular types of social media are more likely to be used by certain demographic and socio-economic profiles for particular purposes; different types of organisation adopt different bridging and engagement strategies with citizens; emergency services show different organisational cultures, receptivity to and use behaviours with regard to social media. This implies that the likely impact of social media will vary depending on the cultural and organisational context in which it is used. The second rival explanation needs to be supported by evidence to show that social media runs as a parallel activity within ES systems, and is utilised by individuals or small groups as a ‘bonus feature’ without being integrated, thereby restricting the capacity of ES to respond more effectively to incidents and leading to citizens not fully understanding the benefits of using social media in emergencies. This evidence could not be fully substantiated – although a pattern repeated across different cases studied showed that the introduction of social media into ES organisational systems and practices can be disruptive, and that some organisations will be resistant to these disruptive changes, leaving social media as an ‘outsider’ or ‘exotic’ feature that is utilised only by social media ‘champion’s and innovators. The evidence to support the third rival explanation – that social media analysis tools contribute to system dysfunctionality – is relatively strong. What seems to be a repeated pattern is that integrating information from social media apps and other sources within ES control systems can disrupt ‘normal’ information flow and lead to system overload. This could lead to the opposite result to that intended – a situation in which situational awareness in ES control rooms is undermined. The analysis highlighted two possible ‘influencing factors’ – defined as ‘contextual conditions that might enable or impede mechanisms’. These are:
104/211 D2.4: Impact of social media, Version V1, PU Cultural contextualisation - familiarity with and routine embedding of social media in individual daily life will increase the probability that citizens will use social media to engage with other citizens and ES in emergencies and will shape their use behaviours Organisational culture - the impact of SM information on ES management of emergencies depends on key organisational culture factors like the efficiency of ES info-management systems; skills base and experience of using social media. The analysis shows that ‘cultural contextualisation’ is likely to have a moderate contributing effect to EmerGent’s expected outcomes. The evidence suggests that EmerGent is most likely to achieve its intended outcome of developing a stronger connection between citizens and emergency services by improving citizens’ awareness of emergency services information needs (both in terms of content and quality) among those already using social media. It also implies that that the strategies ES devise to maximise the potential of social media need to reflect a typology of different cultural protocols. Conversely, ES cannot assume that all citizens uniformly will respond in the same way to the same messages. Another implication is that if the EmerGent guidelines are going to effect change, they have to cater for a range of experiences and needs. Finally, in order to help emergency services use the tools provided by EmerGent they need to be provided with help and guidance on how to do this. Similarly, the analysis shows that ‘organisational culture’ is likely to have a moderate contributing effect to EmerGent’s expected outcomes. The evidence suggests that the use of social media is not yet common practice for all emergency services and they often lack the practice and skills of identifying and interpreting relevant information. This implies that the scale of EmerGent impact in relation to emergency services broadcasting use of social media is likely to vary depending on existing organisational practices. Step 5: Gather new evidence on the implementation of the intervention This stage focused on gathering new evidence to revise and strengthen the contribution story, with particular attention paid to the identified weaknesses in the story. This evidence was gathered and analysed from the following sources: The evaluation results of the EmerGent ICT system (the web interface for emergency services – ESI - and a mobile App for citizens). This covered the second iteration of testing, following an initial ‘scenario-based’ evaluation with representatives of emergency services, which led to the re-design of the ESI interface. Five of the eight planned evaluations in this second iteration were included in our analysis (three remaining evaluations are scheduled): a 2-day test of the App at the OSCE event in December 2016; a 6-day live demo (paper-based) in Poland in January 2017; a 1-day live demo in Salzburg in February 2017; a 4 week field trial in Dortmund Fire Service in March-April 2017 and a 4 week field trial in Hamburg Fire Service in April-May 2017. The evaluation results were published in D.6.4: User-oriented Evaluation. A total of 13 on line survey responses were collected for the ‘App’ evaluation. However, only 6 of these were ‘analysable’. A total of 13 interviews were completed for the remaining four evaluations combined. The four-week field trials in Dortmund and Hamburg were also the subject of additional research carried out by the Tavistock Institute (outlined below). Wave 3 of the EmerGent case studies, carried out by the Tavistock Institute. This involved three groups of case studies: i) eight examples of the use of particular
105/211 D2.4: Impact of social media, Version V1, PU technologies to collect and analyse social media data in organisations, including emergency services (Eindhoven, Rotterdam, the National Police Service Centre in the Netherlands, Antwerp, the National Crisis Centre of Belgium, Greater Manchester Police Service, Frankfurt Fire Service, Graz Emergency Service) ii) examples of ‘terrorist-related’ emergencies – the Boston Marathon attack and the attack on Brussels Airport iii) case studies of field trials of the Emergent ESI and App in Dortmund and Hamburg Fire Departments. The methodology involved interviews with the key actors involved in setting up and implementing the trial. These two case studies were supplemented by a case study of the live demo in Salzburg, which involved observation and interviews with participants. Surveys of Emergency Services, carried out by the Tavistock Institute. A ‘baseline’ survey was conducted in October 2014, based on an Opportunity sample of 696 survey responses from emergency service staff and other users across 27 European countries. A follow up survey was conducted between April and June 2017 and received 473 survey responses across 25 European countries. Surveys of Citizens - representative online surveys of citizens were conducted in four European countries: Germany, Italy, Netherlands and the UK. A total of 7,071 citizens of these four countries completed the survey. The survey explored citizens’ attitudes towards the use of social media in emergency situations and to what extent that have already used social media in emergencies. This stage of the contribution analysis therefore includes evidence on both ‘observed’ and ‘expected’ outcomes. Following the approach used in Step 3, the following Tables show the evidence collected and analysed for each of the primary explanatory mechanisms specified in the Theory of Change, covering i) the evidence that supports the explanatory mechanism ii) the evidence that refutes the explanatory mechanism, together with estimations of the strength of this evidence. The Tables show the evidence and analysis carried out in Step 3 together with the additional evidence and analysis carried out in Step 5.
106/211 D2.4: Impact of social media, Version V1, PU Table 15: Evidence collected – Mechanism 1
Mechanism 1: ‘Awareness-Reach-Trust’: Citizens and ES increased awareness of each other’s activities, behaviours and information needs improves the quality of information shared; increases ES use of social media before, during and after an emergency; reduces dependence on traditional media and reaches a larger number of people; reduces negative impact of misinformation and establishes trust in social media
Supporting Evidence Strength
London riots supported co-operation between citizens and ES through discouraging some Rather strong people from participating, organizing citizens in cleaning up the damage, raising money to compensate those negatively affected
Flooding case studies show evidence of social media contributing to ES-citizen co-operation on informing, alerts, response co-ordination. Rather strong
Flooding case studies show that social media are efficient and effective in broadcasting messages to particular groups of citizens and the media, even more so if a sufficient number Rather strong of followers have been accumulated before the emergency and these followers contain citizen amplifiers.
Evidence from the ESI field trials and demos highlighted several benefits for decision-making, incident reporting or informing the population. Rather strong
Case studies of the Dortmund and Hamburg trials showed that the EmerGent system is Moderately useful to observe reactions and feedback from citizens. strong
Three of the case studies implemented in the third wave (Eindhoven, Antwerp and Moderately Rotterdam) use a co-production model to engage with citizens in all phases of the emergency strong management cycle cultivating close relationships with individuals and groups. This enhances the impact of social media in emergencies and establishes the municipalities, and their public safety and emergency services, as credible and trusted sources of information for citizens and other stakeholders.
The third wave of case studies shows that social media data can make a distinction in an Moderately emergency situation between people who are actually affected; people who think they are strong affected and people who are not directly affected.
The surveys of emergency services staff show that the use of social media has increased in Moderately five key areas: receiving messages from public; gaining situational awareness; two-way strong communication with public; sharing information on behaviour in emergencies and sharing information on how to avoid accidents. They showed that in 2017 88% of staff thought social
107/211 D2.4: Impact of social media, Version V1, PU media could be a useful tool for emergency services to share information with citizens and 73% agreed it is important for emergency services to use social media to keep in touch with the public during emergencies. – both up by 5% since 2014.
The Citizen Survey showed that almost half (45%) of citizens across the four countries have used social media during an emergency.
Refuting Evidence Strength
The flooding case studies show that many emergency services did not have the structures, Rather strong processes, skills and resources to use information from social media effectively during the floods. Moderately The field trials and demos were carried out with a very small number of respondents strong
The third wave of case studies shows variability in expertise and experience across agencies Moderately and departments, leading to some inconsistency in the quality and usefulness of the social strong media analysis carried out, and highlighting the need for greater standardization in recruitment and training and the application of some kind of quality assurance system for social media analysis.
The Citizen Survey showed that most citizens don’t share information because of issues Moderately around credibility of social media content; technical concerns; data privacy concerns; strong concerns about false rumours.
108/211 D2.4: Impact of social media, Version V1, PU Table 16: Evidence collected – Mechanism 2
Mechanism 2: ‘Process-Effectiveness-Embedding’. ES integrate social media into their workflows or create new processes to do so, which allows them to respond more effectively to incidents and leads to better understanding by citizens of the benefits of using social media in emergencies, so that their use becomes standard practice, and citizens share information and act as first responders.
Supporting Evidence Strength
After the London riots, the Met Police put in place a number of organisational and practical Rather weak changes in how they used social media in daily operations
The flooding case studies show in several of the cases that ES learning about needs and Rather strong familiarity with relevant technologies prompted a shift in how they view and / or use social media during and after each emergency, so that using social media to broadcast information has become part of the emergency management routine.
Evidence from field trials and demos shows that overall, most participants emphasized a Rather weak positive attitude towards the system. Participants highlighted the ease of use and simplicity of the ESI.
Field trial Participants indicated that it was a good opportunity to get a general situational overview of local events and developments in social media, but also to get more specific Rather weak information sometimes e.g. to prepare for or predict emergencies.
Case studies of the Dortmund and Hamburg trials showed that the EmerGent system is Moderately useful for getting an overview of what’s going on. This is very helpful and relieves the strong pressure of looking at several platforms for relevant information at the same time.
The third wave case study analysis showed that almost all case studies had staff with a role Moderately relating to social media, though it is comparatively rare to find roles that deal exclusively strong with managing and monitoring social media
Refuting Evidence Strength
The flooding case studies show that many emergency services did not have the structures, Rather strong processes, skills and resources to use information from social media effectively during the floods.
The third case study wave identified a number of factors that inhibit strong connections
109/211 D2.4: Impact of social media, Version V1, PU between social media and ES, including resistance to use social media (info) by frontline or Moderately dispatch staff; organisational concerns about loss of control and reputation; Insufficient staff strong for social media ambitions; fast pace of technological change; variable social media expertise between departments and agencies can create inconsistencies of data analysis.
The 2017 Survey of ES staff showed that the take-up of particular tools (such as the EmerGent system or Coosto) is likely to be greatest among, so called, ‘early adopters’ or Moderately ‘converts’. The survey showed that only 37% of respondents often use social media to share strong information with the public about how to avoid accidents or emergencies. Less than half utilized social media often for all five categories of use covered in the survey, particularly during an emergency.
110/211 D2.4: Impact of social media, Version V1, PU Table 17: Evidence collected – Mechanism 3
Mechanism 3: ‘Workflow-Functionality-Awareness’. By integrating social media into the ES workflow, social media data are used more often during emergencies and more information is posted by ES; functionality for assessing a situation is better integrated and situational awareness improves.
Supporting Evidence Strength
None of the emergency services in the flooding case studies have had technologies in place None to fully harness the information generated in social media during emergencies.
Case studies of the Hamburg and Dortmund field trials showed that the EmerGent system is Moderately stable, and that the layout is clear and core functions self-explanatory. The fact that it is a strong web-based system has advantages for organisations operating under strict IT security protocols and hence may restrict very tightly what software can be downloaded onto work computers. The geo-tagging functionality makes it superior to other (free) solutions (e.g. Tweetdeck) and other than google alerts, the system provides real-time information.
A result of the case study from the Hamburg trial was that those involved saw it as a good Rather weak way to experiment more with social media (analysis) technology as the organisation as a whole is still not very advanced in this regard.
The third wave case study analysis identified a range of examples of social media-ES system integration. These examples show that emergency services, municipalities and other Moderately agencies responsible for public safety use a combination of free and paid-for tools to access strong and analyse social media.
The third wave of case studies showed that in most case studies, the use of social media is Rather strong integrated into the emergency management cycle and that use of social media is an integral / routine part of their work and is supported by senior management.
The 2017 Survey of emergency services showed that 78% agreed Social media could be Moderately useful for gaining situational awareness information during emergencies. – up by 10% from strong 2014.
Refuting Evidence Strength
Flooding case studies showed instances where social media information overload Moderately contributed to reduced functionality and integration of ES IT systems strong
The London riots case shows that turning information into intelligence is more challenging Moderately during extreme events resulting in millions of tweets and other social media data in a short strong period of time.
No data on citizens’ use of the EmerGent App were included in demo and field trial Rather weak evaluations. Evaluation of the App was carried out by ES staff – only 6 survey responses were
111/211 D2.4: Impact of social media, Version V1, PU
analysable, and of these the use of the app during the field trials in Dortmund and Hamburg was very low. Rather weak Evidence from field trials and demos shows that participants found it challenging to achieve
a wide distribution of the EmerGent App among citizens and perceived the processing of individual app messages too time-consuming.
The Citizen Survey shows that the use of apps of relevance during emergencies is still in its Moderately infancy. 28% of citizens in the Netherlands downloaded emergency apps; only 7% of citizens strong in the UK, and 16% in Germany and in Italy had done so.
The next set of Tables present the revised ‘Alternative primary explanatory mechanisms’ together with the ‘Influencing Factors’ identified using the Relevant Explanation Finder (REF) methodology. The tables presented are shortened, with duplicate information – including the data from ‘Step 3’, presented above - removed; please find the full tables in Appendix 10.
112/211 D2.4: Impact of social media, Version V1, PU Table 18: Alternative Primary Explanation Mechanisms M1A (2)
Alternative M1A: Citizens and ES less strongly connected through social media Explanation
Identifiers The third wave of case studies and the Citizens survey showed the main disadvantage of relying heavily on Twitter and Facebook as sources of social media information and channels of communication is that they ignore other social media platforms growing in popularity with particular demographics (like Snapchat and Instagram) and increasingly have less access to certain groups like young people (who tend to use closed source channels like WhatsApp). This leads to fragmentation of channels of communication between different citizen groups and ES.
However, the third wave of case studies show that the impact of social media in emergencies is enhanced when social media is incorporated within broader systems and processes aimed at supporting transparency and open government. In Eindhoven and Rotterdam, social media has contributed to establishing a culture of ‘information open- ness’ which in turn has laid the foundations for establishing the municipalities, and their public safety and emergency services, as credible and trusted sources of information for citizens and other stakeholders. This has consequently increased the motivation of citizens and other stakeholders to play an active role as information providers and validators. Moreover, the widespread engagement of citizens has reduced the need to cultivate ‘citizen communities’ as part of the emergency management system.
Implication The evidence implies that the likely impact of social media will vary depending on the cultural and organisational context in which it is used.
The evidence shows that ES and EmerGent both need to take account of changes in social media technologies, which reflect greater differentiation in the use of different services by different groups for different purposes.
113/211 D2.4: Impact of social media, Version V1, PU Table 19: Alternative Primary Explanation Mechanisms M2A (2)
Alternative M2A: Parallel lines Explanation
Identifiers The Wave 3 case studies show a more broader, and more sophisticated, adoption of social media tools than was indicated in previous case study waves, with several examples of ES incorporating data management tools – like Coosto, Tweetdeck and Hoostsuite - within their existing systems. These tools are used in different ways, according to the priorities, target groups and strategies of the organisation, and are contextualised for purpose. In most case studies, the use of social media is integrated into the emergency management cycle and that use of social media is an integral / routine part of their work and is supported by senior management, or at a higher political level.
Implication The wave 3 case studies showed clearly how the organisational culture of a service will shape how social media tools are embedded in the organisations’ emergency management processes / procedures. The case studies suggested a typology of ‘whole system enthusiasts’, ‘bottom up enthusiasts’, ‘top down professionals’ and ‘worried pragmatists’. EmerGent needs to ensure that the EIS, App and Guidelines reflect this typology.
114/211 D2.4: Impact of social media, Version V1, PU Table 20: Alternative Primary Explanation Mechanisms M3A (2)
Alternative M3A: Social media analysis tools contribute to system dysfunctionality Explanation
Identifiers The wave 3 case studies showed that information overload per se is not the major impediment to improved situational awareness. Two major problems encountered by ES in handling social media information were, first, identifying the geographical location of messages and, second, false positives - despite the use of complex search strings and other functions, most tools still provide a lot of information that is not relevant to a particular emergency or service; this means that during a major incident those using such a tool may struggle to identify relevant messages or resort to searching by hashtags (#) only which can result in missing other important information not shared in this way.
Another consistent theme across all of the case studies in wave 3 was the role traditional media – and journalists – play in emergency situations and the importance of integrating this human element into systems aimed at improving situational awareness.
Implication However, wave 3 of the case studies highlighted two key issues that the EmerGent interface would need to address in order to improve situational awareness – identifying the geographical source of information, and reducing the amount of irrelevant information gathered. The EmerGent system needs to be able to handle search strings to strike a balance between general categories (e.g. capturing information about ‘fires’ ‘floods’ etc.) but also to reflect local contextual conditions, particularly the geography of the area in which the tools are being used.
The case study analysis shows that integration of social apps, IT systems and ES IT systems needs to also involve integration of social media with traditional media – and in particular creating the conditions whereby social media experts work with journalists.
115/211 D2.4: Impact of social media, Version V1, PU Table 21: Influencing Factor – Cultural Contextualisation (2)
Influencing Factor Cultural contextualisation
Identifiers The Wave 3 case studies – particularly in Rotterdam and Eindhoven – showed that when social media is embedded as routine practice as part of an ‘open government’ approach it increases the engagement of citizens as co-producers of information in emergencies.
Implication The evidence suggests that EmerGent is most likely to achieve its intended outcome of developing a stronger connection between citizens and emergency services by improving citizens’ awareness of emergency services information needs (both in terms of content and quality) among those already using social media. It also implies that that the strategies ES devise to maximise the potential of social media need to reflect a typology of different cultural protocols. Conversely, ES cannot assume that all citizens uniformly will respond in the same way to the same messages. Another implication is that if the EmerGent guidelines are going to effect change, they have to cater for a range of experiences and needs. Finally, in order to help emergency services use the tools provided by EmerGent they need to be provided with help and guidance on how to do this.
116/211 D2.4: Impact of social media, Version V1, PU Table 22: Influencing Factor – Organisational Culture (2)
Influencing Factor Organisational culture
Identifiers Third wave case studies show a key training need now in emergency services and similar agencies is to keep up with trends in social media.
The ES Surveys showed that 86% of staff thought that acquiring skills to use social media analysis tools was very important.
The case studies of the field trials showed that, what functionalities are needed and how the IES tool would be used depends on the target group.
The ES Surveys showed that 84% of staff thought that organisational culture was important in ensuring social media was used effectively.
Both Dortmund and Hamburg trials suggest that introducing the system into an emergency services organisation is likely to require some training.
Both Dortmund and Hamburg trials suggest that there is likely to be a need for local adaptation of the EmerGent suite of tools so that usage is in line with organisational processes and practices.
The case studies of the field trials show that keywords need to be adjusted (continuously) to a particular context
Implication The evidence suggests that the use of social media is not yet common practice for all emergency services and they often lack the practice and skills of identifying and interpreting relevant information. This implies that the scale of EmerGent impact in relation to emergency services broadcasting use of social media is likely to vary depending on existing organisational practices. One of the key challenges emergency services have at present is accessing and making sense for response purposes of the increased volume of social media information that is generated during emergencies. In none of our case studies did emergency services have a way to systematically access social media data. It would therefore seem that here lies a great potential impact for EmerGent.
Step 6: Revised Contribution Story This step entailed producing a revised contribution story, using the evidence assembled and analysed in Step 5. Strengths of the Contribution Story EmerGent’s first expected outcome – that Citizens and ES are more strongly connected through social media in emergency situations – appears to be supported by the evidence. The contribution analysis provides four ‘rather strong’ items of supporting evidence; four ‘moderately strong’ items and no ‘weak’ items. This evidence supports the explanatory mechanism that citizens and ES increased awareness of each other’s activities, behaviours and information needs improves the quality of information shared; increases ES use of social
117/211 D2.4: Impact of social media, Version V1, PU media before, during and after an emergency; reduces dependence on traditional media and reaches a larger number of people; reduces negative impact of misinformation and establishes trust in social media. The London Riots and flooding case studies provide strong evidence that demonstrate co-operation between citizens and ES in terms of providing information and reassurance; disseminating alerts; organizing groups of citizens to support the ES response. The evidence shows that social media are efficient and effective in broadcasting messages to particular groups of citizens and the media. The case studies of the field trials of the EmerGent interface and App carried out in Dortmund and Hamburg showed that the EmerGent system is useful to observe reactions and feedback from citizens. The third wave of case studies provide clear evidence that social media is a credible and trusted source of information for citizens and other stakeholders. The surveys of emergency services staff show that the use of social media has increased in five key areas: receiving messages from public; gaining situational awareness; two-way communication with public; sharing information on behaviour in emergencies and sharing information on how to avoid accidents. The Citizen Survey showed that almost half (45%) of citizens across the four countries have used social media during an emergency. EmerGent’s second expected outcome – that social media tools are better connected to the EMC – requires evidence to support an assumed mechanism in which ES integrate social media into their workflows or create new processes to do so, which allows them to respond more effectively to incidents and leads to better understanding by citizens of the benefits of using social media in emergencies, so that their use becomes standard practice, and citizens share information and act as first responders. The evidence to support this explanation is mixed (one ‘rather strong’ item; two ‘moderately strong’; three ‘rather weak’). The London riots case study showed that, after the riots, the Met Police put in place a number of organisational and practical changes in how they used social media in daily operations. Similarly, the flooding case studies show in several of the cases that ES learning about needs and familiarity with relevant technologies prompted a shift in how they view and / or use social media during and after each emergency. Evidence from field trials and demos of the EmerGent interface and App shows that overall, most participants emphasized a positive attitude towards the system. Participants highlighted the ease of use and simplicity of the ESI. Additional case studies of the Dortmund and Hamburg trials showed that the EmerGent system is useful for getting an overview of what’s going on. The third wave case study analysis showed that almost all case studies had staff with a role relating to social media, though it is comparatively rare to find roles that deal exclusively with managing and monitoring social media. EmerGent’s third expected outcome – that Social apps, the EmerGent IT system and ES IT systems are better or more strongly integrated – requires evidence to support an assumed mechanism in which, by integrating social media into the ES workflow, social media data are used more often during emergencies and more information is posted by ES; functionality for assessing a situation is better integrated and situational awareness improves. The initial contribution analysis found no evidence to support this explanation. However, the second round of analysis identified relatively strong evidence to support this explanation (one ‘rather strong’ item; three ‘moderately strong’ items and one ‘rather weak’ item). Case studies of the Hamburg and Dortmund field trials showed that the EmerGent system is stable, and that the layout is clear and core functions self-explanatory. The fact that it is a web-based system has advantages for organisations operating under strict IT security protocols. The third wave case
118/211 D2.4: Impact of social media, Version V1, PU study analysis identified a range of examples of social media-ES system integration. The 2017 Survey of emergency services showed that 78% agreed Social media could be useful for gaining situational awareness information during emergencies – up by 10% from 2014. Weaknesses of the Contribution Story The counter-evidence against EmerGent’s first expected outcome – that Citizens and ES are more strongly connected through social media in emergency situations – is relatively strong, but limited (three ‘moderately strong’ refuting items and one ‘rather strong’ item). The flooding case studies showed that many emergency services did not have the structures, processes, skills and resources to use information from social media effectively during the floods. The third wave of case studies reinforced this evidence, showing variability in expertise and experience across agencies and departments, leading to some inconsistency in the quality and usefulness of the social media analysis carried out. The Citizen Survey showed that most citizens don’t share information because of issues around credibility of social media content; technical concerns; data privacy concerns; concerns about false rumours. In addition, it should be recognized that the field trials and demos implemented to evaluate the EmerGent interface and App were carried out with a very small number of respondents The evidence also undermines some of the assumptions underlying this explanation. The assumption that ‘emergency services will reach audiences more efficiently using social media’ is set against evidence from the flooding case studies which show that efficient reach comes with the trade- off of receiving greater citizen responses which can place an additional burden on emergency services – the way they respond then affects the mood of citizens. The assumption that ‘citizens are homogenous in their use of social media’ is set against evidence to show that citizens take up different roles during emergencies, and that particular types of social media are more likely to be used by certain demographic and socio-economic profiles for particular purposes. The evidence shows that social media are used for different purposes by different people, and have cultural specificities. The counter-evidence against EmerGent’s second expected outcome - that social media tools are better connected to the EMC – is relatively strong but limited (two ‘moderately strong’ refuting items and one ‘rather strong’ item). The flooding case studies show that many emergency services did not have the structures, processes, skills and resources to use information from social media effectively during the floods. The third case study wave identified a number of factors that inhibit strong connections between social media and ES, including resistance to use social media (info) by frontline or dispatch staff; organisational concerns about loss of control and reputation; Insufficient staff for social media ambitions; fast pace of technological change; variable social media expertise between departments and agencies can create inconsistencies of data analysis. The 2017 Survey of ES staff showed that only a minority of ES use social media often in emergencies, and that take-up of social media tools is likely to be greatest amongst early adopters. The evidence also undermines some of the assumptions underlying this explanation. The assumption that ‘ES are homogenous in their use of social media’ is not supported by the evidence. The evidence shows that emergency services differ in their social media skills and knowledge as well as their organisational context. Organizations vary in the speed and enthusiasm with which they adopt social media. Some are resistant to innovation With regard to EmerGent’s third expected outcome – that Social apps, the EmerGent IT system and ES IT systems are better or more strongly integrated – the contribution analysis
119/211 D2.4: Impact of social media, Version V1, PU highlighted five items of refuting evidence (three ‘moderately strong’ and two ‘rather weak’). The flooding case studies showed instances where social media information overload contributed to reduced functionality and integration of ES IT systems. The London riots case showed that turning information into intelligence is more challenging during extreme events resulting in millions of tweets and other social media data in a short period of time. The demo and field trial evaluations of the EmerGent App involved only six analysable survey responses, and these showed that the use of the App during the field trials in Dortmund and Hamburg was very low. The evidence also undermines some of the assumptions underlying this explanation. The assumption that ‘People will use an EmerGent App’ is not borne out by the evidence. The assumption that ‘Social Media services are stable’ is also not supported. The evidence shows that social media are fragile and their attributes and use patterns change rapidly. Evidence from field trials and demos shows that participants found it challenging to achieve a wide distribution of the EmerGent App among citizens and perceived the processing of individual App messages too time-consuming. The Citizen Survey shows that the use of apps of relevance during emergencies is still in its infancy. 28% of citizens in the Netherlands downloaded emergency apps; only 7% of citizens in the UK, and 16% in Germany and in Italy had done so. Alternative explanations and Influencing factors The contribution analysis identified three possible rival explanatory mechanisms to the three assumed mechanisms embedded in EmerGent’s Theory of Change. For the revised Contribution Story, data on actual observed outcomes were also included. The three possible outcomes identified, and their associated explanatory mechanisms are: Citizens and ES less strongly connected through social media - citizens and ES increased awareness of each other’s’ activities leads to fragmentation of behaviours, sometimes in oppositional ways to those intended. Parallel lines - social media runs as a parallel activity within ES systems, and is utilised by individuals or small groups as a ‘bonus feature’ without being integrated, thereby restricting the capacity of ES to respond more effectively to incidents and leading to citizens not fully understanding the benefits of using social media in emergencies. Social media analysis tools contribute to system dysfunctionality – the added flow of information overloads ES IT systems and hampers functionality and situational awareness is subsequently decreased. In the case of the first rival explanation, the evidence that the use of social media weakens the connection between citizens and emergency services is weak. However, the evidence does suggest that neither citizens nor emergency services are homogenous groups. Citizens take up different roles during emergencies; particular types of social media are more likely to be used by certain demographic and socio-economic profiles for particular purposes; different types of organisation adopt different bridging and engagement strategies with citizens; emergency services show different organisational cultures, receptivity to and use behaviours with regard to social media. This implies that the likely impact of social media will vary depending on the cultural and organisational context in which it is used. The evidence shows that ES and EmerGent both need to take account of changes in social media technologies, which reflect
120/211 D2.4: Impact of social media, Version V1, PU greater differentiation in the use of different services by different groups for different purposes. The second rival explanation needs to be supported by evidence to show that social media runs as a parallel activity within ES systems, and is utilised by individuals or small groups as a ‘bonus feature’ without being integrated, thereby restricting the capacity of ES to respond more effectively to incidents and leading to citizens not fully understanding the benefits of using social media in emergencies. This evidence could not be fully substantiated – although a pattern repeated across different cases studied showed that the introduction of social media into ES organisational systems and practices can be disruptive, and that some organisations will be resistant to these disruptive changes, leaving social media as an ‘outsider’ or ‘exotic’ feature that is utilised only by social media ‘champion’s and innovators. The wave 3 case studies showed clearly how the organisational culture of a service will shape how social media tools are embedded in the organisations’ emergency management processes / procedures. The case studies suggested a typology of ‘whole system enthusiasts’, ‘bottom up enthusiasts’, ‘top down professionals’ and ‘worried pragmatists’. EmerGent needs to ensure that the EIS, App and Guidelines reflect this typology. The evidence to support the third rival explanation – that social media analysis tools contribute to system dysfunctionality – is moderate. What seems to be a repeated pattern is that integrating information from social media apps and other sources within ES control systems can disrupt ‘normal’ information flow and lead to system overload. This could lead to the opposite result to that intended – a situation in which situational awareness in ES control rooms is undermined. Wave 3 of the case studies highlighted two key issues that the EmerGent interface would need to address in order to improve situational awareness – identifying the geographical source of information, and reducing the amount of irrelevant information gathered. The EmerGent system needs to be able to handle search strings to strike a balance between general categories (e.g. capturing information about ‘fires’ ‘floods’ etc.) but also to reflect local contextual conditions, particularly the geography of the area in which the tools are being used. The case study analysis shows that integration of social apps, IT systems and ES IT systems needs to also involve integration of social media with traditional media – and in particular creating the conditions whereby social media experts work with journalists. The analysis highlighted two possible ‘influencing factors’ – defined as ‘contextual conditions that might enable or impede mechanisms’. These are: Cultural contextualisation - familiarity with and routine embedding of social media in individual daily life will increase the probability that citizens will use social media to engage with other citizens and ES in emergencies and will shape their use behaviours Organisational culture - the impact of SM information on ES management of emergencies depends on key organisational culture factors like the efficiency of ES info-management systems; skills base and experience of using social media. The analysis shows that ‘cultural contextualisation’ is likely to have a moderate contributing effect to EmerGent’s expected outcomes. The evidence suggests that EmerGent is most likely to achieve its intended outcome of developing a stronger connection between citizens and emergency services by improving citizens’ awareness of emergency services information needs (both in terms of content and quality) among those already using social media. It also implies that that the strategies ES devise to maximise the potential of social media need to
121/211 D2.4: Impact of social media, Version V1, PU reflect a typology of different cultural protocols. Conversely, ES cannot assume that all citizens uniformly will respond in the same way to the same messages. Another implication is that if the EmerGent guidelines are going to effect change, they have to cater for a range of experiences and needs. Finally, in order to help emergency services use the tools provided by EmerGent they need to be provided with help and guidance on how to do this. Similarly, the analysis shows that ‘organisational culture’ is likely to have a significant contributing effect to EmerGent’s expected outcomes. The evidence suggests that the use of social media is not yet common practice for all emergency services and they often lack the practice and skills of identifying and interpreting relevant information. This implies that the scale of EmerGent impact in relation to emergency services broadcasting use of social media is likely to vary depending on existing organisational practices. One of the key challenges emergency services have at present is accessing and making sense for response purposes of the increased volume of social media information that is generated during emergencies. In none of our case studies did emergency services have a way to systematically access social media data. It would therefore seem that here lies a great potential impact for EmerGent. Conclusions and Implications As noted in the Introduction, ultimately, Contribution Analysis aims to infer ‘plausible association’ between a project and a set of relevant outcomes by means of systematic inquiry. To demonstrate this ‘plausible association’, EmerGent’s embedded Theory of Change needs to meet the following five criteria Mayne [Mayn11]: Plausibility: Is the theory of change plausible? Implementation according to plan: Has the program been implemented with high fidelity? Evidentiary confirmation of key elements: To what extent are the key elements of the theory of change confirmed by new or existing evidence? Identification and examination of other influencing factors: To what extent have other influencing factors been identified and accounted for? Disproof of alternative explanations: To what extent have the most relevant alternative explanations been disproved? The Table below summarises the key findings of the Contribution Analysis in relation to these criteria. Table 23: Key findings of the Contribution Analysis Criteria Result Justification
Plausibility Fairly plausible The contribution analysis shows that the primary explanatory mechanisms embedded in EmerGent’s Theory of Change are largely supported by the evidence. This evidence suggests that the activities carried out in the project are likely to lead to its desired outcomes and goals. However, all of the assumptions embedded in the Theory of Change can be questioned to some degree
Implementation Fairly high The project has been implemented according to the DoW and the according to plan project work plan. All of the outputs specified have been delivered. However, evidence of actual outcomes of the project outputs is limited.
122/211 D2.4: Impact of social media, Version V1, PU
Evidentiary To some The contribution analysis shows that all three primary explanatory confirmation of extent mechanisms embedded in the EmerGent Theory of Change show key elements stronger levels of supporting evidence compared with refuting evidence (ratios of 2:1 for mechanism 1; 1.4:1 for mechanism 2 and 1.3:1 for mechanism 3). However, the analysis highlights some weaknesses in the Theory of Change that would need to be addressed with regard to future exploitation and sustainability of EmerGent’s key outputs.
Identification and Identified but Two influencing factors were identified: cultural contextualization and examination of not fully organisational culture. Both these factors are likely to have a moderate other influencing accounted for to high influence on EmerGent’s expected outcomes. They highlight factors specific implications for the future exploitation and sustainability of EmerGent’s key outputs.
Disproof of Three Three ‘direct rival’ explanatory mechanisms were identified: alternative alternative Citizens and ES increased awareness of each other’s’ activities explanations explanations Identified - 2 leads to fragmentation of behaviours out of 3 can be Social media runs as a parallel activity within ES systems discounted. Added flow of information overloads ES IT systems and hampers functionality.
For alternative 1, the evidence is weak. For alternative 2, the evidence is weak. For alternative 3, the evidence is moderate.
The contribution analysis shows EmerGent’s Theory of Change is coherent, and relatively plausible, though all of the assumptions embedded in the Theory of Change can be questioned to some degree. The primary explanatory mechanisms embedded in EmerGent’s Theory of Change are largely supported by the evidence. This evidence suggests that the activities carried out in the project are likely to lead to its desired outcomes and goals. However, evidence of actual outcomes of the project outputs is limited, and the contribution story we have assembled is largely based on an assessment of the ‘likely performance’ of the project, and the probability that its expected outcomes are likely to be realized sometime in the future, rather than an assessment of the actual outcomes observed. This further implies that – although the project has been implemented with relatively high fidelity, in terms of meeting the objectives and targets specified in its work plan – EmerGent has some way to go in terms of the ‘distance travelled’ towards its longer-term goals and desired impacts of enhancing the safety and security of citizens before, during and after an emergency; improving situational awareness during an emergency and, ultimately, reducing damage and loss of life. To achieve these desired impacts, the contribution analysis highlights some implications for future exploitation and sustainability of EmerGent: the likely impact of the EmerGent tools will vary depending on the cultural and organisational context in which they are used. EmerGent therefore needs to take account of changes in social media technologies, which reflect greater differentiation in the use of different services by different groups for different purposes
123/211 D2.4: Impact of social media, Version V1, PU the organisational culture of a service will shape how social media tools are embedded in the organisation’s emergency management processes / procedures, and hence the impact EmerGent tools are likely to have. EmerGent needs to ensure that the EIS, App and Guidelines reflect the typology of ‘whole system enthusiasts’, ‘bottom up enthusiasts’, ‘top down professionals’ and ‘worried pragmatists’ identified by the impacts work carried out to improve situational awareness the EmerGent tools need to be able to identify the geographical source of information, and reduce the amount of irrelevant information gathered. The EmerGent system needs to be able to handle search strings to strike a balance between general categories (e.g. capturing information about ‘fires’ ‘floods’ etc.) but also to reflect local contextual conditions, particularly the geography of the area in which the tools are being used given emergency services’ apparent lack of support for an EmerGent ‘App’, more work needs to be done to identify the potential value of an App for emergency services and citizens, and how this value could be realized more attention needs to be paid to how the EmerGent tools can be used to help integrate social media and traditional media within the systems of emergency services if the EmerGent guidelines are going to effect change, they have to cater for a range of experiences and needs in order to help emergency services use the tools provided by EmerGent they need to be provided with help and guidance on how to do this, including advice and support to acquire the specialist skills needed to use the tools.
124/211 D2.4: Impact of social media, Version V1, PU 9 Summary This Deliverable develops and applies the concept of impact assessment resting on a Theory of Change model initially presented in D2.1, developed in D2.2 via a case study exploration of the London riots, and applied and updated via case studies of floods in six countries in D2.3. This is done by a mixed methodology of both quantitative and qualitative research methods, involving online surveys of citizens and emergency services, case studies of the use of technologies to collect and analyse social media data, case studies of the use of social media during recent terrorist attacks in Brussels and Boston, and three trials of the use of the EmerGent tool by emergency services.
Methodology Section 2 sets out the methodology for the different methods adopted as part of this Deliverable. This includes, first, details of the representative online surveys of citizens conducted in the UK, Germany, Italy, and the Netherlands between October 2016 and March 2017 in order to collect robust data and up-to-date information on citizens’ attitudes towards the use of social media for private use and in emergency situations – involving a total of over 7,000 citizens across these four countries; second, a survey of 473 emergency staff members (these included some who completed a previous survey in 2014 and shared their email addresses) between 24th April and 30th June 2017 in order to assess the extent to which attitudes towards social media have changed over the last three years since the beginning of the study. Third, Section 2.3 outlines the methodology adopted for Phase 3 of the case study methodology, which focuses on learning from real examples of the use of social media by emergency services. We were particularly interested in i) examples of emergency services that had embedded the use of social media in their organisational structures, systems and practice ii) specific platforms and technologies that were being used by emergency services iii) examples of real emergencies that illustrated i) and ii). This was extended by the inclusion of two additional case studies with regard to two high-profile terrorist attacks: the Boston marathon bombing in the USA (2013) and the Brussels airport and metro bombings in Belgium (2016). These were used to explore the impact of social media on how ES respond and react in an emergency (Task 2.2) and how citizens groups or communities are engaged in this response, and what impact this has on them (Task 2.3). Finally, Phase 3 of the case studies was concluded with interviews with, and where appropriate, visits to emergency services trialling the EmerGent tool during demonstrations (Salzburg) or real-life events and every day practice (Dortmund and Hamburg). The case studies were used to explore six key themes developed on the basis of the themes and hypotheses developed through the initial London riots case study and the subsequent flooding case studies. These covered: social media tools, platforms or technologies, organisational structures and facilitators, social media users and uses, information validation, emergency service staff skills and resources, and engaging and/or moderating citizen communities.
Online surveys Section 3 presents the main findings from a comparative analysis of the four representative citizen online surveys and from the follow-up emergency services survey. The citizen surveys showed that almost half (45%) of citizens across the four countries have used social media during an emergency – although there is considerable variation across the four countries. This
125/211 D2.4: Impact of social media, Version V1, PU could be related to the different levels of recent emergencies experienced in these countries – thus, it is hypothesised that the reason for the relatively high reported use of social media in Italy could be related to recent tragic events such as earthquakes, avalanches and floods which may have affected large segments of the population. The surveys also showed that overall the main reasons why some citizens are reluctant to use social media during emergencies relate to their view that there are many false rumours on s social media (63%) and to technical issues of using it (57%). The survey of emergency services staff, on the other hand, highlighted that there appears to have been a significant change in attitude among them with regard to the extent to which they see social media as an important tool in general and for gaining situational awareness during emergencies more specifically since 2014. It shows that their use of social media before and during emergencies has increased, although the change is largest for sharing information with the public – both about how to behave in emergencies and how to avoid accidents.
Organisational case study narratives Section 4 sets out the situation in each of the eight examples studied to illustrate the use of particular existing tools or products (such as Brandwatch or Coosto) and how each of the organisations interact with citizens before, during or after an emergency. The eight cases are Eindhoven Municipality (Netherlands); Rotterdam-Rijnmond Safety Authority (Netherlands), the National Police, Unit East, Police Service Center (Netherlands); Antwerp Fire Service (Belgium); Brussels Crisis Centre (Belgium); Greater Manchester Police (UK); Graz Emergency Services (Austria); Frankfurt Fire Department (Germany).
Key findings and lessons learned Section 5 presents key findings from the eight organisational case studies carried out by partners, drawing out the main messages relating to each of the six key themes and research questions.
Use of social media tools and platforms The case study analysis shows that emergency services, municipalities and other agencies responsible for public safety use a combination of free and paid-for tools to access and analyse social media. Coosto is reported as simple to use, automated and provides comprehensive searching, and has a support service that is seen as very useful and effective. In other cases free data management and analysis tools, such as Tweetdeck or Hootsuite were used, but only ‘on demand’, in emergency situations. One disadvantage of some of the commercial tools such as Coosto in particular is that they have been developed as commercial tools mainly for use in communication and marketing scenarios. This means that they are not designed with the aim of finding appropriate, timely and relevant information for emergency scenarios.
Social media users and uses Social media data is seen as providing useful information relating to three main categories: information, behaviour and sense-making. This information analysis model was used extensively in Brussels, Rotterdam and Eindhoven. By interrogating the information coming in
126/211 D2.4: Impact of social media, Version V1, PU through social media according to these three types, social media analysts can make a distinction in an emergency situation between people who are actually affected; people who think they are affected and people who are not directly affected. Collecting and analysing social media data effectively before, during and after an emergency relies heavily on specialist social media to provide the technical expertise necessary to support emergency services and similar agencies in crisis situations.
Organisational structures and facilitators The impact of social media in emergencies is enhanced when social media is incorporated within broader systems and processes aimed at supporting transparency and open government. On the whole, the case studies show relatively modest organisational resistance to the incorporation of social media into emergency service systems and practices. This is because in all of the cases analysed buy-in and ownership of social media strategy has happened at the top level. Where suspicion of and resistance to the use of social media has been encountered, it has been at grass roots level. This has been effectively addressed through awareness-raising campaigns, education and training programmes and, crucially, demonstrating to front-line staff that social media gets results. The main barriers to the successful implementation of social media in emergency services are: the cost of commercial social media analysis tools difficulties in using the analysis tools in particular in relation to choosing the right search terms to identify relevant information some emergency services staff not fully understanding the ‘world’ of social media, the language used in it and how it operates a residual organisational culture in emergency services that devalues the contribution of ‘non-professionals’ difficulties in keeping in touch with and responding to developments in social media technologies, their applications and their users variability in expertise and experience across agencies and departments, leading to some inconsistency in the quality and usefulness of the social media analysis carried out, and highlighting the need for greater standardization in recruitment and training and the application of some kind of quality assurance system for social media analysis.
Information validation Although all of the cases reported that rumour and misinformation had been experienced when using social media in emergencies, it was felt on the whole to be a relatively minor problem. We found no evidence in the case studies of the use of technical devices and tools to support information validation. Instead, information validation relies largely on human actions.
Emergency service training needs A key training need now in emergency services and similar agencies is to keep up with trends in social media. Currently, staff are trained mainly in using Twitter and Facebook, but it is recognised that these are ‘old technologies’. Young People in particular have migrated to newer platforms like Snapchat and Instagram. Training is therefore needed for emergency staff to firstly acquire the competences necessary to use these tools, but also to acquire the
127/211 D2.4: Impact of social media, Version V1, PU skills to be able to understand the questions, behaviours and emotions of their users, so as to be able to tap into their networks.
Moderating citizen communities The case studies suggest that the current trend is moving away from singling out and working with/moderating particular citizen communities and instead putting resources into developing effective collaborative networks with citizens as a whole. However, citizen communities – and particular individuals who, for example, have a high volume of Twitter followers, still play a role in specific emergencies. Lessons learned from the key findings: the case study analysis highlighted four main lessons that have implications for the ongoing design and exploitation of the EmerGent tools and resources: that there are now many rival tools available to analyse and manage social media data; that such tools are used to provide different types of information to emergency services including situational awareness, behaviour and emotions; training in using and analysing social media needs to be directed at the right staff; and finally that guidelines on using social media need to be tailored to the existing levels and types of social media use among emergency services staff.
Use of social media in response to terrorism Section 6 presents two case studies focussing on the use of social media in response to terrorist attacks in Europe (Brussels) and overseas (Boston, USA), in order to complement the evidence from all the other case studies conducted so far (with the exception of D2.1 which focused on the London riots) which have focussed on natural emergencies and/or focussed on the emergency response in Europe alone. The Bombings in Brussels in March 2016 saw social media play a number of roles in the emergency response. Social media was used both in state communication with the public and by emergency services to better understand the situation and informed their response. Analysis of Twitter data revealed that the early stage after this incident appears to be characterised by a large volume of informational posts with relatively little emotional content. As time passes the volume of such posts diminishes and is replaced by a high volume of tweets with emotional content expressing predominantly sadness and shock as well as support and solidarity to the victims. Analysis of the use of social media during and after the Boston Marathon bombings in April 2013 showed how it played a significant role in the information and communication strategy from the official sources as well as among citizens. As such, it was seen as boosting the resilience-building of a whole community. In this respect, the emotional needs of the wider community, and the use of social media to express the need for, and provide, support, meant that this case study is an example of a supportive social media intervention involving both citizens and official organisations. Regarding the quality of the information, this case study provides good evidence of the corrective facility of social media in correcting false rumours and also highlighted the relevance of coordinated communication strategies in emergencies, across different channels and across different organisations working during the emergency.
128/211 D2.4: Impact of social media, Version V1, PU Case study of the EmerGent system Section 7 presents evaluation evidence from two field trials of the EmerGent system at Dortmund and Hamburg Fire Services, carried out in the first half of 2017. The overall question we sought to answer was: “what is the current and potential impact of the EmerGent system?” The case study highlighted four key learning points from the trials, including that keywords need to be adjusted (continuously) to a particular context for the EmerGent tool to work in different situations and context; that what functionalities are needed and how such tool would be used depends on the target group – communication and dispatch teams with emergency services have very different needs for information and the timing of information provided; third, both trials suggest that introducing the system into an emergency services organisation is likely to require some training, and finally, there is likely to be a need for local adaptation of the EmerGent system so that usage is in line with organisational processes and practices. There was agreement in both case study sites that the EmerGent system has potential for future take-up. However, this exploitation was seen to be somewhat dependent on ongoing development of particular aspects of the system to make it more user-friendly.
Conclusions Section 8 presents the main conclusions from this Deliverable based on contribution analysis of the different aims and objectives of the project and the methods used to assess its impact. The contribution analysis shows EmerGent’s Theory of Change is coherent, and relatively plausible, though all of the assumptions embedded in the Theory of Change can be questioned to some degree. The primary explanatory mechanisms embedded in EmerGent’s Theory of Change are largely supported by the evidence. This evidence suggests that the activities carried out in the project are likely to lead to its desired outcomes and goals, although the contribution analysis highlights some implications for future exploitation and sustainability of EmerGent.
129/211 D2.4: Impact of social media, Version V1, PU References [Alap16] Alaphilippe, A. (2016) #BrusselsAttack #Openhouse: the case for social media: Digital Maniak. [online], Available: http://www.digitalmaniak.fr/2016/03/brussels-attacks-the-case-for-social- media/ [AsLe10] Astbury, B. and Leeuw, F. (2010) Unpacking black boxes: mechanisms and theory building in evaluation. American Journal of Evaluation, 31(3), 363–81 [ASW17] Akhgar, B., Staniforth, A. and Waddington, D (2017) Application of Social Media in Crisis Management. Springer. [BBS14] Befani,B., Barnett, C. and Stern, E. (2014) Rethinking Impact Evaluation for Development. IDS Bulletin, 45(6) [BeMa14] Befani, B. and Mayne, J. (2014) Process Tracing and Contribution Analysis: A Combined Approach to Generative Causal Inference for Impact Evaluation. IDS Bulletin, 45(6), 17–36 [BLRK15] Brenner, P.S., LeBlanc, J.L., Roman, A.M. & Kwate, N.O.A (2015) Safety and Solidarity After the Boston Marathon Bombing: A Comparison of Three Diverse Boston Neighborhoods. Sociological Forum, 30(1), 40-61 [online], Available: doi: 10.1111/socf.12144 [BDMR16] Browning, J., Dyer, N., McMahon., H. and Regan, A. (2016) Geospatial Information in Emergency Management – Integrating and disseminating spatial data in first response: The Firm. [online], Available: http://ggs685.pbworks.com/w/file/fetch/107793818/Edited_FIRM.pdf [CCMC13] Cassa, A., Chunara, R., Mandl, K. and Brownstein,J. (2013) Twitter as a Sentinel in Emergency Situations: Lessons from the Boston Marathon Explosions. PLoS Currents, 2(5)
[CSJD+14] Cullen, J., Spielhofer, T., Junge, K., Drabble, D., Ludwig, T, and Reuter, C. (2014), Deliverable 2.1: Concept for Impact Assessment. Emergency Management in Social Media Generation (EmerGent).
[DeTo12] Delahais, T. and Toulemonde, J. (2012) Applying contribution analysis: Lessons from five years of practice. Evaluation, 18(3), 281-293 [GaKa16] Galunic, C. and Kaligotla, C. (2015) The spread of false ideas through social media. INSEAD Working Paper Series 2015/53/OBH [Garb15] Garber, L (2015) Have We Learned a Lesson? The Boston Marathon Bombings and Information Sharing. Administrative Law Review, 67(1) [HKK14] Hu, Q., Knox, C.C. and Kapucu, N. (2014) What Have We Learned since September 11, 2001? A Network Study of the Boston Marathon Bombings Response. Public Administration Review, November/December 2014, [online], Available: http://onlinelibrary.wiley.com.remote.baruch.cuny.edu/doi/10.1111/puar.122 84/pdf
130/211 D2.4: Impact of social media, Version V1, PU [ICT16] International Institute for Counter Terrorism (2016) The Brussels Attacks – 22/03/2016 What do we know? & Insights from ICT Experts. [online], Available: https://www.ict.org.il/Article/1645/The-Brussels-Attacks [JSGC+14] Junge, K., Spielhofer, T., Gieve, M., Cullen, J. and Drabble, D. (2014), Deliverable 2.2: Impact of social media on Emergency Services and Citizens. Emergency Management in Social Media Generation (EmerGent).
[Kemp17] Kemp, S. (2017) Digital in 2017 - Global Overview: We are Social. [online], Available: https://wearesocial.com/special-reports/digital-in-2017-global- overview [LBND12] Lemire, S., Bohni Nielsen, S., and Dybdal, L. (2012) Making contribution analysis work: A practical framework for handling influencing factors and alternative explanations. Evaluation, 18(3), 294–309. [LoRo12] Loi, M. and Rodrigues, M. (2012) A note on the impact evaluation of public policies: the counterfactual analysis. European Commission: JRC Scientific and Policy Reports [online], Available: http://publications.jrc.ec.europa.eu/repository/bitstream/JRC74778/lbna25519 enn.pdf [Mayn11] Mayne, J. (2011) Addressing Cause and Effect in Simple and Complex Settings through Contribution Analysis. In Evaluating the Complex, R. Schwartz, K. Forss, and M. Marra (Eds.), Transaction Publishers. [Mayn12] Mayne, J. (2012) Contribution analysis: Coming of age? Evaluation, 18(3), 270– 280 [Morr16] Moretto, A. (2016) Presumed Guilty: An Examination of the Media's Prejudicial Effect on the Boston Marathon Bombing Trial. New England Journal on Criminal & Civil Confinement, 42(1), 65-100 [MoTu13] Mohammad, S., and Turney, P.D. (2013), Crowdsourcing a word–emotion association lexicon. Computational Intelligence, Volume 29 (3), 436-465. [NeSt17] Neuger, J. and Steams, J., (2017) Belgium’s Worst Terror Attack Ever Leaves 31 Dead in Brussels: Bloomberg Business. [online], Available: "Belgium's Worst Terror Attack Ever Leaves 31 Dead in Brussels". [Pete16] Peterson, A. (2016) Belgian officials urge people to use social media in wake of bombings: The Washington Post. [online], Available: https://www.washingtonpost.com/news/the-switch/wp/2016/03/22/belgian- officials-urge-people-to-turn-to-social-media-in-wake-of- bombings/?utm_term=.649d168f0f6a [SMOA+14] Starbird, K., Maddock, J., Orand, M., Achterman, P., & Mason, R. M. (2014). Rumors, False Flags, and Digital Vigilantes: Misinformation on Twitter after the 2013 Boston Marathon Bombing. iConference 2014 Proceedings , 654–662. [Spie17] Spielhofer, T., (2017) Open doors and fake news - how people used social media after the Manchester attack: Prospect Magazine. [online], Available:
131/211 D2.4: Impact of social media, Version V1, PU https://www.prospectmagazine.co.uk/other/open-doors-and-fake-news-how- people-used-social-media-after-the-manchester-attack [SSMF+12] Stern, E., Stame, N., Mayne, J., Forss, K., Davies, R. and Befani, B. Broadening the range of Designs and Methods for Impact Evaluation (2012) Working Paper 38 [SGSL+15] Sutton, J., Gibson, C.B., Spiro, E.S., League, C., Fitzhugh, S.M.& Butts, C.T. (2015) What it Takes to Get Passed On: Message Content, Style, and Structure as Predictors of Retransmission in the Boston Marathon Bombing Response. PLOS ONE 10(8) [online], Available: doi:10.1371/journal.pone.0134452 [Toul10] Toulemonde, J. (2010) Evaluating impact through a contribution analysis. Workshop at the Prague Conference of the European Evaluation Society. [Vand16] Vanderbiest, N. (2016) Brussels Attacks – Crossover between research and reality: European Emergency Number Association Case Study. [online], Available: http://www.eena.org/download.asp?item_id=215
[Yin02] Robert K. Yin. (2002) Case Study Research. Design and Methods: Applied social research method series, Volume 5. California: Sage Publications
[WWW01] Netherlands Central Bureau of Statistics, http://www.cbs.nl/en-gb [WWW02] UK Office for National Statistics, http://www.ons.gov.uk/ [WWW03] UN Data, United Nations Satistics Division, http://data.un.org [WWW04] Italian National Institute for Statistics, http://www.istat.it/en [WWW05] A language and environment for statistical computing [online] http://www.R- project.org [WWW06] Next Generation Databases, http://nosql-database.org [WWW07] Police Procurement North West Portal, https://www.bluelight.gov.uk/portal/nwpolice/portal.nsf/vLiveDocs/PD-DEVA- 82TLAK?OpenDocument [WWW08] Graz Tourist Site, http://www.graz.at/cms/beitrag/10034466/605976 [WWW09] How Cheshire Fire & Rescue Service use social media to inform and engage with the public [online], Available: content.crowdcontrolhq.com/cheshire_fire_rescue_services_social_media_ma nagement_crowdcontrolhq [WWW10] How Wiltshire Police use social media to provide immediate responses to the public [online], Available: http://content.crowdcontrolhq.com/wiltshire_police_social_media_manageme nt_crowdcontrolhq [WWW11] Key figures for the Brussels-Capital Region [online], Available: h http://statistics.brussels/figures/key-figures-for-the-region#.WXIORYTysdW
[WWW12] Brussels Population (2017) [online], Available: http://worldpopulationreview.com/world-cities/brussels-population/
132/211 D2.4: Impact of social media, Version V1, PU [WWW13] Brussels attack: Zaventem and Maelbeek bombs kill many: BBC News. [online], Available: "Brussels attacks: Zaventem and Maelbeek bombs kill many". [WWW14] Brussels explosions - What we know about airport and metro attacks: BBC News [online], Available: http://www.bbc.co.uk/news/world-europe-35869985 [WWW15] In Brussels Bombing Plot, a Trail of Dots Not Connected: New York Times. [online], Available: "In Brussels Bombing Plot, a Trail of Dots Not Connected". [WWW16] Brussels attacks: Man charged with terrorist offences: BBC News. [online], Available: "Brussels attacks: Man charged with terrorist offences". [WWW17] Brussels airport delays reopening, Belgium lowers attacks toll to 32: AFP.com. [online], Available: "Brussels airport delays reopening, Belgium lowers attacks toll to 32" [WWW18] Twitter Sentiment Analysis Training Corpus (Dataset) [online], Available: http://thinknook.com/twitter-sentiment-analysis-training-corpus-dataset-2012- 09-22/ [WWW19] https://en.wikipedia.org/wiki/Boston [WWW20] https://en.wikipedia.org/wiki/Boston_Marathon [WWW21] https://www.facebook.com/StayStrongBostonStrong/ [WWW22] http://www.staystrongbostonstrong.org/
133/211 D2.3: Impact of social media, Version V1, PU
Appendix 1: Eindhoven Municipality (Case Study Report) Emergent Case Studies: Round 3 Background and Context Eindhoven is a city in southern Holland, a key centre of the Dutch electronics and high-tech sector. The metropolitan area consists of 419,045 inhabitants and the city region has a population of 753,426 (Figure 1).
Eindhoven generates €24 billion of GDP and €55 billion in exports, one-quarter of the Dutch total. It absorbs 36% of all private Dutch R&D spending and is home to globally recognized companies including Philips, the healthcare, lighting and consumer product giant, and ASML, maker of photolithography equipment for the production of silicon chips. 18% of all Dutch automotive jobs are in Eindhoven, and 9% of all life technology employment. The Eindhoven University of Technology, with more than 7,000 students, is one of the top three research universities in Europe and its High Tech Campus founded by Philips houses over 80 companies employing another 7,000 residents
Figure 1: Eindhoven Location Given its prominence as a high-tech hub, it comes as no surprise that Eindhoven Municipality has been a leading exponent of the ‘smart city’ movement and was an early pioneer in finding ways of applying information – especially ‘big data’ – to solve a range of problems, from traffic grid-lock, through developing the ‘circular economy’, to smart policing. Recent thinking around the ‘smart city’ has seen a shift in focus - from developing technologies that impose ‘order from above’ on citizens to create the perfect urban environment - to engaging with citizens as ‘co-producers’ of the knowledge required to make cities become the setting for a smart society. This has led Eindhoven to explore how social media can make a contribution. In line with other Dutch cities – like Rotterdam – Eindhoven’s interest in social media reflects a broader objective of working with ways of delivering ‘open government’. Social media tools, platforms and technologies What types of social media tools, platforms or technologies are being used by emergency services Eindhoven Municipality began exploring how to utilize social media information to deliver services across a range of areas, including public safety, crisis management and communications, in 2010, with a single Twitter account for a particular project – a service to provide information on road works and traffic problems (later entered for the ‘Galjaard Prize’ - a prize awarded annually to a successful government communication project). Following the success of this initial pilot, the Municipality appointed two full-time communication co-ordinators to work with social media. In 2013, the department of Communication and the Customer Contact Center collaborated to set up a ‘web care team’ to monitor and respond to communications on Twitter and Facebook. To handle the volume of
134/211 D2.3: Impact of social media, Version V1, PU traffic, they later invested in ‘Hootsuite’ – a social media analysis platform that is primarily used by organisations for marketing, customer engagement and analysis of communications campaigns. The Municipality then switched to the ‘Coosto’ platform to monitor, access and analyse social media information, because they thought Coosto provided data search functions that were simpler to use, more comprehensive and fully automated. Coosto is used for a wide range of purposes – not just emergencies. For example, recently there was a major power supply failure. Coosto was used to monitor social media messages to identify citizens’ concerns about what was happening and when supply would be resumed. Another example is ‘Kings Day’ – the most important holiday in the season and the day when Eindhoven is reputed to host ‘the biggest party in the Netherlands’: “Not only does the town use crisis monitoring during events, but on a daily basis Eindhoven monitors sentiment as well. We’d just started using social media proactively, when we noticed a video of major abuse in the party area going viral on social media. We had to decide how to communicate this public affair. Our analysis based on Coosto information helped reach decisions for our communication strategy”. On Kings Day, the Department of Communication works with the police to monitor social media traffic so that the mood of the party can be monitored, potential incidents identified and any incidents that then occur assessed on actions that need to be taken (Figure 2): “First, we set the GPS and GEO filter. Then, we monitor any peaks or drops in sentiment and activity in order to spot a potential crisis. If there are peaks or drops, we use the trend analysis and the maps to reach decisions.”
Figure 2: Sentiment and Activity Monitoring in Eindhoven on Kings Day The Department of Communication monitors social media traffic 24/7. In parallel, the police also use Coosto to monitor social media information. When there is an emergency, or ‘special situation’, a situation team is set up to co-ordinate the response. This includes representatives of the other 20 municipalities with whom the city of Eindhoven works; the police, fire and other services, depending on the situation. Most of the social media information fed into Coosto comes from Twitter. As is the case with the Rotterdam Safety Authority, Eindhoven Department of Communication works closely with traditional media – particularly the local newspaper, which is mainly published on-line, to channel and reinforce its messages to citizens and to help verify information.
135/211 D2.3: Impact of social media, Version V1, PU What are seen as the strengths and weaknesses of these? Coosto was selected following a sales pitch by three suppliers. Although the decision to choose Coosto was based on its technical functionalities and cost, an added bonus was that Coosto was developed in Eindhoven and its offices are in the city. Eindhoven Department of Communication report they are ‘very happy with it’ and have experienced few problems using it. The only major problem cited is that sometimes the search terms used do not produce the expected results. When this happens, Coosto provides support to rectify the problem. The level and quality of the support service provided was cited as a major benefit of using Coosto. Twitter is the preferred source for social media information because it is an open system and responds very quickly to developing situations. The main issue using Twitter information is that good search parameters are needed to extract the relevant information from irrelevant tweets. This means that the analysts who use Coosto to monitor and analyse social media information have to be highly skilled and trained in their task. What are the organisational resistance dynamics and barriers to the use of social media and how were these addressed? The organisational framework that governs the way social media is used in Eindhoven is relatively flexible. Different agencies have their own systems – for example the police have their own version of Coosto and their own Twitter account. As noted above, in emergency or ‘special situations’, a situation team is convened comprising representatives of key stakeholders from the city and 20 additional municipalities in the region. The decision-making structure of this team is quite hierarchical. To carry out actions based on analysis of social media data, the analysis team working on the ground – which will typically include a small team of 3 – 5 people from the Department of Communications – need to get approval from the high level decision makers from the various branches represented, like police, fire services and politicians. This can sometimes lead to delays in taking the necessary action. There is no evidence of a culture of resistance to the use of social media within Eindhoven municipality or its emergency services. This, as noted above, reflects Eindhoven’s position as a leading, progressive high-tech hub and as a municipality committed to the ‘smart city’ movement and to open and transparent government. It was also suggested that this area of the Netherlands politically has a history of communitarianism, exemplified by its tendency to form ‘left-wing coalitions’ in government – which reflects a broader culture of collaboration and ‘people working together’. Social media users and uses Which types of staff or other stakeholders are most likely to use social media during an emergency? As outlined above, the City of Eindhoven routinely monitors social media information on a daily basis. This is co-ordinated by a small team of 3-5 communication officers from the Department of Communication, who work closely with emergency services, statutory agencies and local journalists. All of these stakeholders are represented in emergencies via the ‘situation team’ that is convened when a crisis occurs. Most of the emergency situations dealt with involve fires – although there are occasional incidents at Eindhoven airport mainly involving aircraft with engine or landing gear problems. The Department of Communication has built up a library of search terms for use with Coosto. This includes search terms for ‘special subjects’, for example individual political leaders (Aldermen). A core base of around 40 search strings is routinely used. What types of information are emergency services most interested in? In emergency situations, searching and analysis of social media information focuses on three things: i) information – is there a lack of information outside? Are people asking particular questions? Ii) behaviour – are people moving towards the area where a disaster is happening and do they need to be
136/211 D2.3: Impact of social media, Version V1, PU warned? Iii) emotions – what feelings and sentiments are people expressing about a situation? Do they need to be reassured? What kinds of social media citizen roles would be most useful for emergency services? As in the case of Rotterdam-Rinjmond Safety Authority, the city of Eindhoven has worked to develop an ‘open government’ environment in which all citizens are engaged as co-producers of information and knowledge. This means that 95% of the social media information processed through Coosto comes from the general public. However, the key ‘influencers and movers’ within the community are well- known and in certain situations their activities on social media are monitored individually. For example, there was a recent situation in Eindhoven in which around 700 refugees needed to be accommodated. Social media was used in this situation to set up local area committees, with the support of key influencers and players, to handle the situation. Information validation How can emergency services be persuaded that information provided via social media from citizens is credible and trustworthy? As in the case of the Rotterdam Safety Authority, at the heart of information validation is the City’s investment in open government. The logic is that open government will help to build up trust and credibility over time with citizens; they will then be more motivated to provide social media information that is accurate. The Department of Communication reports that false rumours and misinformation are not often encountered in the social media traffic monitored and analysed. What procedures and tools are used to validate such information efficiently and effectively? Information validation is done through triangulation. In emergency situations, members of the situation team – which includes police, the fire department and health agencies – compare and cross-check the information in a Report (every 2 hours) provided by the analysts from the Department of Communication. Prior to this, the individual members of the analysts team also cross-check information internally. The team also scans local media stories as another point of comparison and validation. How can reliability and accuracy be supported, technically? The city of Eindhoven doesn’t use automated data analysis to validate social media information. As outlined above, information validation is done through triangulation. Emergency service staff skills and resources What skills are required to help emergency service staff to find, interpret and make use of social media data? The most important skill is understanding social media and its language. Second is the capacity to differentiate – to be able to know what’s important and not important. Another important skill is how to interpret the emotions and sentiments underlying the information people provide on social media. This requires a high level of emotional intelligence – particularly empathy. Finally, in the Eindhoven context, diplomacy is important – or more specifically social media analysts need to know how to provide advice, based on social media analysis, to key decision-makers in the situation team, in a way that will optimize the chances of the advice being acted on. What guidance or training would be most useful for them? Eindhoven Department of Communication provides training in using Coosto for its own staff, as well as staff from Eindhoven’s emergency services and for staff from the 20 municipalities the city collaborates with. The training programme focuses firstly on the technical skills necessary to operate Coosto – particularly the skills needed to apply search terms effectively, but also on the ‘soft skills’ required to
137/211 D2.3: Impact of social media, Version V1, PU interpret and use information – particularly understanding the language of social media and the colloquial terms used, and knowing the local geography and culture: “The best social media analysts are those who like solving puzzles and people who can solve things through trial and error. Also, you have to think like the man in the street. You need to know their slang and have their local knowledge”. The main skills gap identified is variability in expertise and experience across agencies and departments. This leads to some inconsistency in the quality and usefulness of the social media analysis carried out in the city as a whole. This suggests a need for greater standardization in recruitment and training and the development and application of some kind of quality assurance system for social media analysis. Moderating citizen communities As noted above all citizens in Eindhoven are in principle engaged as co-producers of information and knowledge. However, the key ‘influencers and movers’ within the community are well-known and in certain situations their activities on social media are monitored individually and sometimes operate within special committees that are set up to respond to a particular emergency situation. Key Conclusions Eindhoven Municipality is another good example from the Netherlands of the successful embedding of social media within the strategic and operational culture of a city and its emergency services. A key factor contributing to this has been the policy of ‘open and transparent government’ adopted by the city authorities (as is also the case in Rotterdam). This has helped establish the Municipality as a credible and trusted source of information for citizens and other stakeholders and, in turn, has increased the motivation of citizens and other stakeholders to play an active role as information providers. The open government policy is in turn linked to Eindhoven’s longstanding role in the ‘smart cities’ movement. Eindhoven has long been interested in using information – particularly ‘big data’ – to solve urban problems. The recent shift in focus in the smart city movement from ‘top- down’ solutions to collaborating with grass roots organisations and citizens has favoured the use of social media information as part of the city’s crisis management strategy. The widespread engagement of citizens has reduced the need to cultivate ‘citizen communities’ as part of the emergency management system. Social media analysis in emergency situations is co-ordinated by a ‘situations team’ composed of representatives of the 21 municipalities in the region, emergency services staff and other statutory agencies. There is little evidence of a culture of resistance to the use of social media in emergencies. Coosto is used to access, extract and analyse social media data. This is because the system is seen as simple to use, provides comprehensive searching and is fully automated, and because the support service provided by Coosto is seen as very useful and effective. Twitter is used most extensively to engage with citizens and other stakeholders, because of its open-ness and rapid response capabilities. The main skills required to use social media information effectively are being able to understand social media and its language, particularly local language; the capacity to differentiate between important and unimportant information; the ability to interpret the emotions and sentiments underlying the information people provide on social media; the ability to convert social media analysis results into credible action plans that decision-makers will respond to. The main obstacle to successful use of social media information is variability in expertise and experience across agencies and departments, leading to some inconsistency in the quality and usefulness of the social media analysis carried out, and highlighting the need for greater
138/211 D2.3: Impact of social media, Version V1, PU standardization in recruitment and training and the application of some kind of quality assurance system for social media analysis. Themes and Questions for Case Studies Round 4 The usefulness of the EmerGent data mining tools – what value added can a data mining application produced by EmerGent contribute to emergency services who use Coosto? Do different types of organisational structure imply differences in the effectiveness of the results of social media analysis? Do hierarchical structures reduce the impact of social media information analysis? Emergency service staff training and guidance – what kind of pedagogic approach and learning content is required to support emergency staff in acquiring the competences necessary to become fluent in the ‘language of social media’ and in acquiring the necessary emotional intelligence? Environmental assessment – is the ‘information-behaviour-emotions’ model used by Eindhoven generalizable across different emergency services and situations? What contribution can the ‘smart city’ approach make to the more widespread and effective use of co-produced information by emergency services and statutory agencies?
139/211 D2.3: Impact of social media, Version V1, PU Appendix 2: Rotterdam-Rijnmond Safety Authority (Case Study Report) EmerGent Case Studies: Round 3 Background and Context The Rotterdam-Rijnmond Safety Authority is a governmental organisation which carries out duties on behalf of the region’s municipalities in the areas of disaster relief, crisis management, risk management, fire services, ambulance and medical assistance (Figure 1). Rotterdam-Rijnmond is one of 25 safety regions in the Netherlands.
A safety region is defined as ‘a geographical area within which governing bodies and operational services co-operate in terms of fire prevention, calamity response, crisis control and medical assistance in order to maintain and increase the safety of the inhabitants’. Its objectives are:
To improve protection for inhabitants against safety risks To improve the help offered during, and after incidents, hazards and crises To improve fire-prevention, fire- repression, medical assistance at accidents and disasters, risk management and centralised regional crisis management To increase overall managerial and operational performance
Figure 1: The Rotterdam-Rijnmond Safety Region The Rotterdam-Rijnmond Safety Authority covers a large and complex environment, including urban areas with 1.2 million inhabitants, a diverse infrastructure with a global harbour and 32.000 sea-going vessels, 110,000 inland vessels and around 800.000 shipping movements. This complexity is highlighted by the fact that the Safety Authority has over 80 Contingency Plans in place for various emergencies. 1 Reflecting this complexity and diversity, another distinguishing feature of the Safety Authority is its large network structure and organisational depth. As Figure 2 shows, the Safety Authority is a multi- stakeholder collaborative structure, involving 15municipalities in the Rijnmond area; emergency services; other public agencies (Regional public broadcasting: – RTV Rijnmond, Public Transport – RET, Ministry of Defence, Water Boards, Royal Dutch Rescue Service, Health Services) and commercial organisations, including energy companies and 60 companies in the harbour and industrial areas in cooperation with the municipality of Rotterdam. 2 Co-ordination and strategic governance is provided through a ‘Common Regulation’ that incorporates four administrative bodies: Governing Board Executive Board
1 Source: ‘Working for Safety’. Presentation, Rotterdam-Rinjhold Safety Authority, 2013. 2 Source: op. cit.
140/211 D2.3: Impact of social media, Version V1, PU SRR: aSafety la Regionrge President ne twork Safety directorate.
Figure 2: Structure of the Rotterdam-Rijnmond Safety Authority The Authority operates from a central office overlooking the port of Rotterdam, where it also co- ordinates a General Emergency Control Room for Police, Fire brigade and Ambulance Services (112). In ‘crisis mode’, a ‘crisis management team’ from within the Safety Region will co-ordinate the response to an emergency. It should be borne in mind that around 1,500 emergencies are logged every week within the Safety Authority region and only 1 or 2 of these need to be handled by the crisis management team. Social media tools, platforms and technologies What types of social media tools, platforms or technologies are being used by emergency services Social media is extensively used by the Safety Authority as a key tool within its overall crisis management strategy. This is based on an ‘environmental analysis’ model that has three inter-related dimensions (or ‘topics’): Assessing information needs Assessing the requirements for damage reduction Assessing ‘meaning’ (or ‘emotions’). The model is applied iteratively in a crisis situation at regular intervals to re-appraise the ongoing situation (every 10, 30 or 60 minutes depending on the nature of the emergency). To support this model, the Authority uses social media, amongst other sources, in two main ways: first, to analyse social media messages from within the region to identify incidents and take action on the analysed information. second, to actively communicate with citizens and other stakeholders. Social media is an important channel in how the Authority informs citizens. Analysis of social media messages is done using Coosto. Coosto is a corruption of the name of the famous French underwater explorer, chosen, according to its manufacturer, because the product
141/211 D2.3: Impact of social media, Version V1, PU enables users to “Just like him …dive deeper to discover more. Not in the deep sea, but in the social web”. Coosto is a broad-brush tool with a wide range of applications, including supporting customer services management, publishing and assessing the impact of social media content and supporting ‘reputation management’, including analysis of competitors and markets. However, most emergency services that have started to use Coosto do so because of its data research capabilities. This enables the capture, filtering and analysis of content from social networking sites, blogs, forums, (online and offline) news sites, radio and television. It provides Boolean search operators and cross-reference searching; filtering; geolocation analysis and data export. Coosto is used by the Rotterdam Safety Authority primarily for real-time situation analysis: "We use Coosto on a structural basis to monitor situations. As a public assistant we deal with fires, accidents or other incidents on a daily basis. We focus a lot on prevention, but some situations we cannot prevent. That’s why we have a real-time dashboard with live information. Social media monitoring functions as an important signal for us”. 1 Although the Safety Authority monitors social media information on a daily basis, it should be stressed that the main objective of doing this is to look for tendencies and trends, rather than focus on single incidents. As the Safety Authority’s Communication Advisor observes: “We monitor on a daily basis, but the effectiveness regarding single incidents is slim to none. For single incidents, we have our 112 emergency response center. People give us a call in case of an emergency”. 2 In this context, Coosto is used by the Rotterdam Safety Authority for three main reasons: its capacity to search, capture and analyse large amounts of information; its capability of doing this in real-time and the ability to create a ‘dashboard’ for sharing information with colleagues and other stakeholders. “In case of an incident, I set up a search query in Coosto based on words that describe the situation and based on location. If there is a fire on the Erasmus bridge for instance, I search for messages in which “fire” or “smoke” and “Rotterdam” or “Erasmusbrug” are mentioned, but I also search for all messages sent from within a specific radius of the Erasmus bridge. This means that I can also collect messages not mentioning fire, but which could also contain important information. For me, the key features of Coosto are unlimited search, real-time monitoring and real-time analytics of social media messages." 3 To achieve its second objective in using social media – to inform and alert – the Safety Authority built a platform - rijnmondveilig.nl – that combines Twitter, Facebook, email and SMS to inform the media and the public and to alert them to disasters and incidents on a 24-hour basis. For any given incident, the platform initiates a live blog that follows the incident from start to end and which feeds into Twitter and other social media (Figure 3). As an indication of the scale, coverage and utilization of the platform, during 2015 the platform delivered 208 live blogs and 1,365 tweets, with 3,524 retweets or replies, which were seen 31,068,714 times. It posted 196 messages on Facebook, with 2851 comments or shares and published 100 ‘Risk Tips’. The platform has three main goals: To ensure that emergency services are trustworthy and trusted, and that they ‘get their facts right’ To communicate incidents quickly
1 Source: https://www.coosto.com/en/resources/user-success-stories/regional-analysis-at-a-safety-authority 2 Interview data 3 Source: https://www.coosto.com/en/resources/user-success-stories/regional-analysis-at-a-safety-authority
142/211 D2.3: Impact of social media, Version V1, PU To support transparency and open-ness by government agencies: “As a government we have no secrets. Everyone can know everything”.
Figure 3: rijnmondveilig.nl - the main social media platform used by Rotterdam Safety Authority What are seen as the strengths and weaknesses of these? Twitter is the most used social media tool by the Authority, and is seen as the most valuable, because it is open (public) and because it supports ‘rapid delivery and response’. This rapid responsiveness has significantly shaped the Authority’s social media strategy – because it became clear to them that, to a large extent, the emergency services and public authorities are often ‘behind the game’ in emergency situations: “Often the first pictures are already on Twitter, before the rescuers are on site. As a government, you are always one step behind. The outside world knows a lot more - and faster - than we do”. This has shaped the Authority’s strategy in two main ways. First, because Twitter provides a potentially high volume, accessible and rapidly available source of information that can be analysed and applied to understand emergency situations, react to them and inform the public about what is happening, it was selected as the main social media tool. Second, because information about incidents is often circulated publicly in a very short time, the Authority has highlighted information validation – particularly responding to rumour and false information – as a key priority in how it works with social media. According to the Authority, the main strengths of the social media used are: They provide a rich source of information to assess the situation at an early stage – what is the situation on the ground? can we see people who are injured? is there a lot of smoke around? They enable emergency services to effectively utilize citizens as a resource to add value to the information the services have, and fill in gaps They contribute to open government. Sometimes, government agencies appear to adopt a stance that they know better than everyone else. This is sometimes reinforced by a tendency to be secretive or uninformative. The Safety Authority takes the line that if you don’t know something, admit it. This attitude, they believe, makes it more likely that citizens will contribute to more effective crisis management by being more motivated to supply useful information. The main weaknesses identified are: Some emergency services staff do not fully understand the ‘world’ of social media, and how it operates
143/211 D2.3: Impact of social media, Version V1, PU Linked to this is a need for staff to undergo extensive training regarding searching and interpreting the social media networks, for example by using search terms in Coosto that aren’t jargonized and reflect the everyday language used by real people on the ground. Organisational structures and facilitators How is the use and analysis of social media integrated into the emergency service’s organisational structures and systems? Social media is now part of the organisational culture and life of the Safety Authority. It is now standard procedure for emergency services to use social media in crisis situations: “We even have entire firefighting teams that are active on Twitter, Facebook and YouTube. They are more in touch with the citizens, the heart of society”. Since the beginning of 2016, the municipalities within the Safety Authority have explored ways of applying the model and approach within their everyday practice. However, a key feature of the approach used in Rotterdam is that social media is not used in isolation. It is seen as one component part of a holistic system that shapes how the whole network operates. This is co-ordinated by the communication experts within the Crisis Management Team, which involves representatives of the multi-stakeholder partnership. The communication experts are all trained in using social media and in social media analytics. A member of the team is always on duty. A formal social media training programme has been set up to ensure that there is a continuing supply of trained personnel capable of using social media effectively in emergencies. The programme uses a ‘train the trainer’ method, so that graduates from the programme can expand the skills base throughout the Rotterdam Safety Authority through mentoring and ‘buddying’ (learning by doing). What are the organisational resistance dynamics and barriers to the use of social media and how were these addressed? Like all organisations, the Safety Authority encountered some scepticism about the value of social media and its effectiveness, and some resistance to its introduction. This mainly emanated from the grass roots, from front-line staff in emergency services rather than from emergency services management or policy-makers in the municipalities. The main driver of this resistance appears to be an endemic part of the organisational culture within emergency services that takes the perspective of ‘we are the professionals and we know it all’. This attitude meant that some front-line emergency services personnel couldn’t see how anything contributed by non-professionals could be valuable or useful. The strategy adopted by the Safety Authority to counter this element of the organisational culture of the emergency services was to put resources into awareness-raising about social media – with a particular emphasis on its practical value. One of the most effective ways of doing this was to use ‘social media dashboards’ in incidents rooms. A wide-screen TV was taken into the incidents room and hooked up to demonstrate how the social media platform was working in real-time, including how information could be analysed using Coosto. Because emergency services staff could see Tweets, and word clouds that showed real situations unfolding, they could see the value of such information and could see that the public, rather than being ignorant or naïve about emergency situations, instead showed a high level of awareness and insight. Social media users and uses Which types of staff or other stakeholders are most likely to use social media during an emergency? As outlined above, because social media has become part of the routine life of emergency services it is used extensively across the spectrum within the Rotterdam-Rijnmond Safety Authority. Fire stations in the region have set up their own Twitter account and, currently, 20 out of 50 fire stations use it. The main beneficiaries of this are the public who engage with the emergency services in two main ways.
144/211 D2.3: Impact of social media, Version V1, PU First it allows them to become ‘citizen informants’. A recent example was the rebuilding of a new Casino in the region. Residents in the vicinity became concerned that the construction was unsafe and alerted the fire department, which then inspected the building and suspended construction until it was made safe. Another, more frequent, intervention by citizens is alerting the emergency services to situations where emergency exits are being blocked. Second, social media is used to spread the message of crisis prevention – particularly with regard to using smoke detectors in the home. A key consumer of social media information in the Rotterdam region are journalists. This is because the ‘environmental analysis’ model used by the Safety Authority adopts a ‘mediation’ or ‘social gatekeeping’ approach to connect citizens with services. This means that the existing collaborative relationship that has been built in the Rotterdam region between citizens and government is reinforced through links between local government agencies and key stakeholders and opinion-leaders within the different communities. Journalists are seen as a key component of this ‘holistic’ approach – not least because the media they work for have a large reach: “We are also in constant contact with the press. Journalists are not - as some might think - our opponents, but our supporters. We work on the basis of trust. What does the journalist hear? What rumours are live? What do they see on the ground? What do they hear from residents?” Journalists are also important in the information validation process (outlined below) because they act as a point of triangulation for information collected and analysed by the Safety Authority. Information collected from journalists and other partners in the network is published on the live blog, after checking its validity. As highlighted in previous case studies – for example the flooding cases in the Elbe and in Tbilisi – the Rotterdam case underlines a trend towards using social media not in isolation but as part of an integrated package that combines social media with traditional media. What types of information are emergency services most interested in? The Authority is particularly interested in eyewitness accounts, especially when supported by posts of images showing the incident. More generally, when analysing social media information using Coosto, the social media analysts look for three main types of information: questions, behaviour and emotions. This can help the Safety Authority decide which kinds of messages are likely to be most effective in informing and alerting citizens. Emotions suggest that people have heard about something and are worried, and so would benefit from a reassuring message. Questions suggest that people are aware of a situation but are not at the centre of it, and so would welcome a response that is informative. Behaviour-type messages suggest people are at the centre of an incident and need actions to be taken – for example sending a fire tender to the vicinity. In most cases, a combination of the three is crafted into a communication strategy and carried out. What kinds of social media citizen roles would be most useful for emergency services? In the Rotterdam-Rijnmond case, the social media citizen role the Safety Authority wants to encourage is a ‘collaborative’ or ‘co-producer’ role. In contrast to other case studies, for example where emergency services try to encourage the involvement of citizen mediators and specific on-line communities, the emphasis in Rotterdam- Rijnmond is on trying to bring citizens as a whole into the multi-stakeholder partnership and network set up within the region to deliver safety. The ultimate aim is to make interaction between citizens and emergency services part of everyday life. Because the system is based on a holistic approach, engaging all stakeholders within the Safety Region – including citizens and journalists – as co-collaborators in creating, accessing and analysing social media information, the system operates as a ‘multiplier’ for information. In one example, on November 6, 2015, there was a large fire at a car dealership in Vlaardingen. Rijnmondveilig.nl was used to issue an ‘NL-Alert’, warning citizens to keep their windows closed. However, monitoring of tweets showed that most citizens were not interested in safety questions, but in the potential effects the fire could have on
145/211 D2.3: Impact of social media, Version V1, PU school closures. People wanted to know whether, for example, they needed to arrange child care so they could go to work. The Safety Authority then worked with the municipality to access schools, who then used their phone system to circulate information. Information validation How can emergency services be persuaded that information provided via social media from citizens is credible and trustworthy? The Safety Authority has taken a holistic approach to social media information validation. This starts with ensuring that ‘open-ness’ is embedded in the way the Authority operates and the procedures and tools it uses. The rationale is that, if the Authority establishes its reputation as an open government, then it will build up trust and credibility over time with citizens; they will then be more motivated to provide social media information that is accurate, and will be more willing to collaborate with the Authority to counter misinformation and rumour. On the emergency services side, as noted above, the Authority has invested resources in promotion and awareness-raising to persuade emergency service staff of the value and credibility of information coming from the public. The tactic of demonstrating the real-time dashboard live to emergency service staff has contributed significantly to establishing the credibility of citizen information. What procedures and tools are used to validate such information efficiently and effectively? Information validation is done through triangulation, as part of the overall ‘environmental assessment’ approach used, through deploying the resources of the multi-stakeholder network: “Reliability is our USP. We check our facts and are in close contact with our partners”. Information is cross-checked with the 112 information room emergency desk, with emergency service staff operating on the ground at the scene of an incident, with journalists and through feedback from citizens. For example, they ask questions of followers of their Twitter account – ‘Do you smell smoke? Is it bothering you?’. Because the Safety Authority has invested heavily in establishing its credibility and reliability credentials, citizens now are willing to provide information that can be used in cross-checking without being asked, for example through eye-witness accounts, backed up with images: “In addition to checking photos and eyewitness accounts, the system will use specific monitoring questions. If one person asks a question on social media, then that question will go live with several people”. How can reliability and accuracy be supported, technically? The Safety Authority don’t use automated data analysis to validate social media information. As outlined above, information validation is done through triangulation, as part of the overall ‘environmental assessment’ approach used. Example of Information Validation Despite the information validation systems in place, rumours and misinformation are still a problem for the Safety Authority. A good example is an incident that happened four years ago when a rumour spread of a gunman approaching a shopping mall in Zuidplein. The rumour spread quickly, generating half a million messages on Twitter, Facebook and WhatsApp within an hour. One of these tweets showed a photograph of police cars at the scene (Figure 4). The Safety Authority then started the process of information validation. They contacted sources in the police, who reported that the alleged gunman was many miles away from the scene. The Safety Authority established that the tweet showing the photograph was a malicious piece of misinformation, generated by someone editing a photograph taken two weeks previously of police outside the mall taking part in a strike action. They then spread the real story through to journalists, who used their own social media networks to dispel the rumour. This was backed up by messages issued by the Safety Authority through the rijnmondveilig.nl platform.
146/211 D2.3: Impact of social media, Version V1, PU
Figure 4: Rumour in Rotterdam Emergency service staff skills and resources What skills are required to help emergency service staff to find, interpret and make use of social media data? The main skills required by emergency staff to find, interpret and make use of social media data are: An understanding of the different social media tools, how they work in different ways and for what purposes A ‘feel’ for the social media networks – particularly the characteristics and behaviours of the people who use the different platforms and tools Fluency in the ‘language’ of social media – particularly in the use of social media ‘slang’ Technical expertise in using Coosto – particularly competence in applying appropriate search terms. An understanding of how communication works during crisis situations. The Safety Authority currently has sufficient capacity to deliver these skills, largely due to its training programme (described above) which uses a ‘train the trainer’ method, so that graduates from the programme can expand the skills base throughout the Rotterdam Safety Authority through mentoring and ‘buddying’ (learning by doing). However, the Authority is conscious that training needs to be constantly refreshed and updated. What guidance or training would be most useful for them? The key training need now in the Safety Authority is to keep up with trends in social media. Currently, staff are trained mainly in using Twitter and Facebook, but it is recognised that these are ‘old technologies’. Young People in particular have migrated to newer technologies like Snapchat and Instagram. Training is therefore needed for emergency staff to firstly acquire the competences necessary to use these tools, but also to acquire the skills to be able to understand the questions, behaviours and emotions of their users, so as to be able to tap into their networks. Moderating citizen communities As noted above, the Rotterdam-Rinjmond Safety Authority has focused on developing effective collaborative networks with citizens as a whole, rather than nurturing citizen communities. However, one area in which the Authority is experimenting with Facebook Groups is incident prevention – particularly preventing domestic fires. Currently, one of the ways in which this is done is through ‘Peter
147/211 D2.3: Impact of social media, Version V1, PU the Fireman’, who hosts a radio show every two weeks to spread the word about smoke alarms. The experiment currently under way is to supplement this radio show with a ‘Peter the Fireman’ live blog and to evaluate whether this has an additional impact on changing behaviours. Key Conclusions The Rotterdam-Rinjmond Safety Authority is a good example of the successful embedding of social media within the strategic and operational culture of emergency services. A significant contributing factor to this success has been the setting up of a multi-stakeholder partnership and network at the strategic level that involves all stakeholders in the design, implementation and management of a holistic crisis management system. This has supported ownership and buy-in of social media at a high level. The partnership also makes a significant contribution to the efficiency and effectiveness of the social media capture, validation and analysis process. Stakeholders in the partnership play a leading role as co-producers of information as well as in cross-checking its validity. The emphasis on establishing a culture of ‘information open-ness’ in the Safety Authority has laid the foundations for establishing the Safety Authority as a credible and trusted source of information for citizens and other stakeholders and, in turn, has increased the motivation of citizens and other stakeholders to play an active role as information providers and validators. This widespread engagement of citizens has reduced the need to cultivate ‘citizen communities’ as part of the emergency management system. The hesitation of emergency service staff to engaging with social media has been significantly reduced by implementing a programme of awareness-raising. A key contributing factor to this has been the adoption of live demonstrations to emergency services staff of how social media information is used. The decision to use Coosto as the main data analysis tool was taken for three main reasons: unlimited search, real-time monitoring and real-time analytics of social media messages. Twitter is used most extensively to engage with citizens and other stakeholders, because of its open-ness and rapid response capabilities. The main barriers to the successful implementation of social media in emergency services are: emergency services staff not fully understanding the ‘world’ of social media, and how it operates; a tendency for emergency services staff to lack fluency in the ‘language’ of social media; a residual organisational culture in emergency services that devalues the contribution of ‘non-professionals’; keeping in touch with developments in social media technologies, their applications and their users. Themes and Questions for Case Studies Round 4 The usefulness of the EmerGent data mining tools – what value added can a data mining application produced by EmerGent contribute to emergency services who use Coosto? Multi-stakeholder partnerships – what kind of partnerships are likely to support the use of social media by emergency services and what governance and operational structures are needed to make them effective? Emergency service staff training and guidance – what kind of pedagogic approach and learning content is required to support emergency staff in acquiring the competences necessary to become fluent in the ‘language of social media’? How can training programmes be designed to keep up with developments in social media? Environmental assessment – is the ‘information-behaviour-emotions’ model used by Rotterdam-Rinjmond generalizable across different emergency services and situations? Citizen-stakeholder engagement – can social media contribute effectively to emergency management without the involvement of citizen communities? Are ‘gatekeeper’ models (for example engaging journalists to act as bridging points) more effective than citizen communities?
148/211 D2.3: Impact of social media, Version V1, PU Media combinations – how can social media be most effectively combined with other media?
149/211 D2.3: Impact of social media, Version V1, PU Appendix 3: Netherlands, National Police, Unit East, Police Service Center (Case Study Report) EmerGent Case Studies: Round 3 Background and Context The Unit East Police Service Centre of the Dutch National Police, operates within the largest Unit of the Netherlands, (8.299 km2 ) serving 3.143.000 citizens and employing 6800 police officers. The Service Centre of the National Police is accessible to citizens via a direct telephone number (09008844) as an alternative for the emergency number (112) to contact the police for non-emergency information or enquiries. The Service Centre employs 220 agents, and has a staff of 21 members (10 Operational Experts and 11 Seniors) and is located in four locations within the Unit.
Figure 1: Map of regions in the Netherlands – East Region is number 2
Social media tools, platforms and technologies How is social media used on a day-to-day basis? The core business of the Service Centre is the telephone, but this has expanded into other forms of communication, including emails, web-forms and more recently social media, including Twitter and WhatsApp. This includes using social media not just to monitor events, but also to find new ways to
150/211 D2.3: Impact of social media, Version V1, PU contact people. ‘But it’s a long way – a struggle – especially in the police to achieve that. We are not very fast in developing things to do that’. What types of social media tools, platforms or technologies are being used by emergency services? Service Centre staff do not use any particular tools to monitor social media – instead this is done just by looking at messages sent to the Service Centre Twitter account, rather than monitoring social media messages in general: to look at any comments made or questions asked. This is not part of the overall structure of the centre, but relies on particular Service Centre staff with an interest and ability of doing this: ‘It is not really a real role in our business’. This is quite a small task though, as only very few questions or remarks are sent to them every day that they have to reply to or take action on: ‘It’s really not a lot of work’. In the case of a major incident or emergency, they use public platforms such as TweetDeck or Hootsuite to monitor what is going on: to find out the different hashtags being used and to follow them. And any information found in this way is shared with colleagues in the emergency rooms: ‘Have you seen this?’ so that they can take any action if needed. In contrast, the national region police service of the Netherlands currently uses Coosto to monitor social media on a daily basis to check social media threats, to spot any terrorist threats, any talk about the Dutch royal family, and relating to high ranked politicians. The Service Centre also used Twitcident (which is now also known as CrowdSense or PublicSonar) as part of an annual fund-raising event called ‘The Glass House’. This is an event in December every year involving famous DJs to raise money for good causes. Twitcident was used to find out what was going on around a big event like that to monitor social media mainly for sentiment: ‘To try to find out how people are feeling; how many people are gathering and where; what are they doing and saying?’ This was to see whether it made sense to use a tool like this as an early warning system before something happens. This could include for example identifying Tweets about issues with crowds: ‘If someone sends a Tweet that it is very crowded where he is, but everyone is happy and cheerful, then there is no need to do anything. But if someone says: “it’s crowded but I don’t like the atmosphere around here now”, then I would suggest sending people around there to check it out and de-escalate the situation before anything more serious happens’. The Service Centre does not use Twitcident, but there are some emergency rooms in the Netherlands that are using a tool developed from it (CrowdSense or PublicSonar) to monitor social media in this way on a daily basis. This is done 24/7 by staff working within the so called ‘Real-time intelligence centre’ – the role of this centre is to find out any additional situational awareness about a particular incident or situation. So, for example, if they are called out to a particular address, then this centre will look for any available information on the internet, social media, or in past files, about it and send it to those responding to the incident on the scene. This could include information on previous arrests, or photos of the scene, etcetera. What are seen as the strengths and weaknesses of these? Twitter and simple public tools such as TweetDeck and Hootsuite were seen as sufficient for the day-to- day tasks needed in the Service Centre – involving looking at Tweets to the Twitter accounts or following particular hashtags during incidents. The other tools were seen as more relevant for those needing to monitor social media on a daily basis to identify particular threats or incidents. A particular strength of Twitcident is that it allows you to archive any messages that are not relevant and move those that are important to a folder – this means that you can keep the dashboard clear for any new messages and delete or remove any irrelevant ones. So you can mark a message as ‘important’ but it still disappears from the dashboard into a folder for important messages.
151/211 D2.3: Impact of social media, Version V1, PU Organisational structures and facilitators How is the use and analysis of social media integrated into the emergency service’s organisational structures and systems? As noted above, monitoring of social media is not an official part of the overall structure of the centre, but relies on particular Service Centre staff who are experienced and enthusiastic about doing this. In fact, a lot of this is done by one member of staff from home – so for example of the ‘The Glass House’ event he did around 90% of the monitoring in his own time: ‘Because you get used to it – you scan the posts for something interesting. And so you sleep for a few hours and the first thing you do when you wake up is you open Twitcident to see if there is anything new and to keep up if anything is going on’. What are the organisational resistance dynamics and barriers to the use of social media and how were these addressed? There are various reasons for the relatively slow adoption of social media in the police. This includes organisational resistance to social media, in particular in relation to safety: ‘How safe is it, if we as police get in contact with the outside world by using Twitter or Facebook?’ Another major barrier is the utility of using social media: ‘What will I get out of it? Why should I do it?’ And finally manpower is another barrier: ‘Do I have enough manpower to do this?’ and ‘Do I want to do this?’ Some people still think that if people want to contact the police, they should use a telephone. Social media users and uses Which types of staff or other stakeholders are most likely to use social media during an emergency? As outlined above, there is no permanent role in the Service Centre with responsibility for social media. So when those who do it on their own initiative are away or on holiday this is not done: ‘I’m currently on holiday, so is anyone answering Twitter? No’. However, in particular emergency rooms there are people with this responsibility – and the national service is running a pilot to explore a more systematic use of social media to both find information and communicate with citizens: they are monitoring everything that is going on relevant to police business. ‘It’s like Coca Cola is monitoring everything that is going on when Coca Cola is mentioned’. One other region is experimenting with the use of WhatsApp to interact with the public. And they are also using another tool called Obi4wan to monitor social media (see: https://www.obi4wan.com/nl/). A few years ago, it was unthinkable to do something like this – they would say to you: “Why should I do this?” when companies were already doing this for years. Information validation What procedures and tools are used to validate such information efficiently and effectively? Reliability is always something that concerns people in emergency rooms – this means that some people are suspicious of using social media: ‘It is always something that concerns people that work in emergency rooms. And they will say: “If I’m talking to someone on the phone, I can tell if it’s true or not”’. Of course on social media it is easy to make fake notifications. But it is possible to check information about the person that is sending the message: ‘If you read what they are doing normally – are they trustworthy or not? And has the profile just started the day of the incident or does he have lots of friends or followers?’ Similarly, it is fairly easy to check the authenticity of pictures sent via social media by checking information embedded in the picture such as the time-stamp and any geo-tags. So it’s quite easy to validate whether the picture was taken where and on the day it was supposed to be – this was done successfully during the recent Brussels terror attacks where it was shown that pictures posted on social media were from an attack in Russia several years earlier.
152/211 D2.3: Impact of social media, Version V1, PU Emergency service staff skills and resources What skills are required to help emergency service staff to find, interpret and make use of social media data? Lack of skills among staff was more of a problem in the past – but now there are more people in the police who are familiar with social media and use it themselves and can search very well, so it’s not such a problem anymore. Moderating citizen communities Police emergency rooms use an app called ‘Civil watch’ to share and ask for information from the public to send out warnings about particular urgent incidents. This is used, for example, to report a person who is missing or if there has been a robbery in a jewellery store and the robbers are on the run – so they will send a warning to the public and ask them to share any information that they see. This is done both online and via an app – and around two or three messages are sent out every day. So, for example, recently there was an 88 year old missing person warning sent via the app which reached 6000 people in the local region – and he was found.
153/211 D2.3: Impact of social media, Version V1, PU Appendix 4: Antwerp Fire Service (Case Study Report) EmerGent Case Studies: Round 3 Background and Context The Antwerp Fire Service (AFS) or Brandweer Zone Antwerpen covers the city, the port and two communities – with around 600k inhabitants. Antwerp is the second largest port in Europe and is regarded as a high risk area due to the high concentration of petro-chemical facilities; there is a lot of construction and it is generally expanding a lot as a city. There are eight fire stations in Antwerp, including three in the port area and five in the city area. These used to be quite old fashioned, but have been increasingly modernised over the last 15 years, so they have been updated a lot. The AFS employs 700 operational staff who are all full professionals (no volunteers) and 80 support staff (lawyers, communications, finance, HR, etc.). It is an independent service – i.e. it is financed by the city, but it is ‘independent’ of it; however, the mayor is on the board of the fire service, so he is ‘our real boss’ as 80% of funding comes from the city. It deals with around 22,000 interventions per annum – this includes 10% of fires and 15% of floods – the rest is made up all other kinds of technical interventions, including car accidents, supporting the ambulance service, industrial accidents in the port, etc. More recently they have also had to focus on dealing with the aftermath of terrorism. There has been no major attack in Antwerp, although there have been incidents (arrests) and it is regarded as a high risk area as one of the key extremist groups was founded in Antwerp and alongside that Antwerp has a very large Jewish community.
154/211 D2.3: Impact of social media, Version V1, PU Figure 1: Map of Antwerp port area Interviews for this case study were conducted with: Bert Brugghemans: Chief Fire Officer of AFS Bart Bruelemans: Emergency manager of the city of Antwerp/seconded to AFS Jasmien O: AFS spokesperson and social media manager
Social media tools, platforms and technologies What types of social media tools, platforms or technologies are being used by emergency services? In practical terms, they always have software running to collect and analyse social media data – they use Coosto: “This is a good tool, but not perfect – it works quite well but there are issues with it, as it wasn’t designed for emergency services. But the way we’ve configured it makes it work for us. But it’s very difficult to get it right – we do have false positives and things like that, but we’ve just learned to deal with it”. The main way Coosto is used is to set up a very general search for any bits of social media information that could be relevant to an emergency. This is done by using very general search terms such as accident (crash, cars, trucks, dangerous), fire, explosion, flood, storms, and terrorism. This is then combined with Antwerp district or place names. To make sure they do not miss any posts, they also use slang and dialect words to denote particular areas. “We have built up new terms over the years – every time something picks up you add it. You need to think very carefully about search terms. Coosto then collates the info – we use it in dashboards – this links with the search in real time”. They also use Coosto to follow particular accounts on Twitter, such as “the Port of Antwerp” and use filters to exclude some false positives. This includes, for example, mentions of the term ‘Brand’ in messages – in Dutch this means ‘fire’ – in order to exclude any posts written in English talking about particular ‘brands’. Their searches still result in a lot of false positives, particularly in relation to the term ‘fire’: “With Coosto we don’t filter out ‘fire’ too much so we get them from all over - but as this is a serious event we’d rather get too many than too few”. During particular emergencies, such as a storm or port fire, they will use more specific search terms and filters as part of a specific dashboard – this will be opened in a new window alongside the standard search – to ensure that the information received is relevant to the emergency. How is it used on a day-to-day basis? The AFS social media manager runs Coosto on her laptop continuously and will look at it ‘every few minutes’. If she spots several tweets about an event or emergency at the same time, she will check it and pass on any relevant information, including videos or photos, to the dispatch centre. “If we get extra information about an incident we would send the photos to the officer”. EXAMPLE: BUS ON FIRE Bert described the following incident and how social media was used to provide situational awareness: “There was a bus on fire on the high-way and before we’d even got a call about it, we’d received a picture via social media of the bus on fire … and shortly afterwards we got a call about it; so we could use the information from looking at this photo to share key information to the fire service responding to the call and also to inform our general response to the incident: for example, it showed that the bus was on the hard shoulder and not blocking the traffic – it was a very hot day, in the middle of the summer; so if it had been blocking the traffic, this would have meant big tail-backs which could have led to other incidents and response needs e.g. providing water to people stuck in their cars for hours. The photo also showed that the bus was empty – so there were no immediate victims. So even for a relatively small incident like this you can learn an awful lot from social media”.
155/211 D2.3: Impact of social media, Version V1, PU Most of the time Coosto delivers false positives – messages that do not relate to an emergency or that are not located in Antwerp. However, on average, they will spot something about once a week that is relevant and they need to respond to. Most recently, this included a message on Twitter about a suspected fire in a particular neighbourhood: “It was someone who mentioned, who asked if there was a fire in his neighbourhood – he saw a lot of police and fire cars – then we looked at our info and we responded and provided a lot more info about it”. Coosto provides a way therefore to pick up on conversations via social media (not directed at the AFS) and to intervene to provide further information, guidance and to correct false rumours. The Antwerp police service uses the same tool (Coosto) and shares information with them using a shared platform (called Telegram). So, for example, if there is a fire, then they will share information with each other about information on the fire or impacts on traffic: “They deal with the traffic; we deal with the fire”. Overall, social media is used on a day-to-day basis to collect two types of information: a) Big Data: to detect trends and to inform the big picture about an event b) Micro-information: ‘to find the needle in the hay stack’ – i.e. to find individual bits of information, such as videos or photos or descriptions, that can aid the emergency response.
EXAMPLE: FIRE IN PORT On the 1st September 2016, there was an industrial fire in the Antwerp port and social media was once again used to provide information. Bert described how a video on Twitter found using Coosto provided valuable information to warn citizens and help the fire service in their response: “There was a fire in the port and very quickly we had pictures and videos available of it”
156/211 D2.3: Impact of social media, Version V1, PU Figure 2: Video of smoke shared on Twitter This video was again available almost instantly and they could use it to make key decisions and gain valuable information – e.g. about the smoke direction and how high above the ground the smoke was drifting – so they could use this to provide information and advice to citizens about the emergency (e.g. to close their windows etc.). It was also shared with their communications people via a WhatsApp group and with operational staff during the incident via a dedicated ICT system. They use this system to send pictures/videos to operational staff as they are travelling to the incident – this is used to give a better situational overview of the incident before they even arrive at the scene, and allows them to make decisions at an early stage (e.g. about what incident command level is needed). ‘You’re talking about the first 10 minutes of an incident – after that you have helicopters and things like that … but it’s much slower than what you can get via social media’.
What are seen as the strengths and weaknesses of these? Even though social media acts as a valuable source of additional information during emergencies, it is impossible to rely on it as it may not be available for particular incidents or certain parts of the day: ‘Of course you cannot rely on the availability of social media – for example, if the fire happens at 3am in the morning, you won’t get it as you won’t have anyone walking around taking pictures of it. So it’s not reliable in that sense’. As noted above, the AFS identified some limitations of their existing tool (Coosto) to identify relevant information on social media in a timely fashion. One of the main reasons for this was thought to be the fact that Coosto has been mainly developed for commercial use – to monitor customers’ views and interactions about brands – and has not been developed with emergency services in mind. AFS is currently considering using a different, more expensive, social media analysis tool – CrowdSense/PublicSonar (see: http://publicsonar.com/). This tool is thought to have better search algorithms, as well as a database of local street names, and provides social media messages without a delay (for Coosto there is a delay of a few minutes after a message is posted) and has been developed specifically for emergency services (ES). Organisational structures and facilitators How is the use and analysis of social media integrated into the emergency service’s organisational structures and systems? Like most emergency services the AFS started using social media a while back but in quite an ad hoc and low profile way and mainly to share information with the public rather than to collect and analyse data. The first time they saw the power of social media and realised they needed to take it more seriously was during the Love Parade incident in 2010 – as they noticed a disconnect between what the crisis management team were saying and what was available on social media via YouTube. As a result of this, they started experimenting with social media as a service more and more. Another key incident was Pukkelpop 2012 – where a large storm hit during a pop festival, resulting in five deaths. This was the first time in a crisis in Belgium that people used social media at a large scale to share information with each other and to help each other. From them onwards the AFS grew their use of social media – they started with using social media mainly as a communication channel but more and more to look for information before, during and after a crisis. The increased use of social media by the AFS was aided by the fact that Bert B. is very senior within the organisation, and that the development of crisis management procedures happened only relatively recently in the AFS: “If we did this in the 1990’s then social media would not yet have been available – but we developed crisis management between 2005 and 2015 ..”. He also thought that there were
157/211 D2.3: Impact of social media, Version V1, PU other factors which aided the adoption of social media analytics by the AFS: “it takes an open culture in the organisation – to allow influences in. It also takes a bit of luck: who your boss is and who you are”. On a staffing level, they have two permanent social media staff members and one public information officer position (staffed more or less 24/7 by different people – who are on call or in the office via shift work). The role of social media staff members is to work on the website and with the media, but also to mine social media and to use social media for communication purposes. But ‘if an incident happens, they’ll go full-time onto social media, so that if a crisis happens, we will send a tweet out about it within a few minutes’. For example, if there’s a fire, they will send a tweet about the smoke and provide a link to a FAQ on their website that provides advice about what to do in case of smoke: “Social media is used to communicate messages about how to prevent fires beforehand and what to do to get help during the crisis”. Normally, if a citizen asks for assistance via Twitter they do not promise a response – they are advised to use 112 instead. But during a crisis: "we do respond … but we don’t advertise for it, as we’d still prefer them to call in; if they do, we will send them an intervention number so if later they call up, we can link the information they’ve sent by social media and also they can use it to track our response”. The AFS social media channels and website are also used to share information about the emergency response during a crisis and to raise awareness among citizens of the numbers of calls they have to deal with. This is seen as a way of reducing the number of complaints for slower than usual response times during an emergency. During normal operations, the expected response time is 12 minutes – two minutes to get ready and ten to drive to the location. However when there is a major event this causes a waiting time problem with large demand. During a recent flash flood, for example, they received 1500 calls in four hours, which means that waiting times goes up massively and managing resources efficiently becomes paramount. Social media and the website are used to manage expectations in this situation: ‘We also send out links to maps to give an overview of the city and with the interventions we’ve dealt with – very much to raise awareness and understanding among citizens of the magnitude of the incident – so they know what we’re doing and that we’re working hard’. During the recent flooding incident, they only had one complaint afterwards from all these calls: ‘People are happy because they realise what we’re doing and how hard we are working’. They also use SM for prevention. So, for example: ”If we know that Sunday will be thunderstorms in this area, even though we don’t know exactly where, we’ll send out a warning on Friday and Saturday … So we start communication right now – what to do to prepare for it! It’s not easy though, as sometimes in won’t happen – and people are often not interested in it now before the event … but we try to do it”. What are the organisational resistance dynamics and barriers to the use of social media and how were these addressed? There are still people in the AFS who do not trust information from social media – particularly among operational staff members: “At the operational level, there is some scepticism – they are not always aware of the impact of information on their decision making – they don’t realise how important it is”. But at the strategic level most staff members were said to be supportive of the use of social media: “via so many good examples of how we used social media in the past to inform situations. So they know it is important”. This means that the social media team will often target information at senior staff to make sure it is passed on and taken seriously by the operational teams: “If we need them to see it we go to their senior officers and they send it on and they have to take notice as it come from higher up the command”. There is also some resistance to the use of social media to find out about emergencies among dispatch centre staff. “They are often resistant to information from social media – they don’t see it as clear and structured enough – they just want people to call them on 112 and tell them what’s happening – but society is changing and people are less and less willing to use the phone and so we need to change with them. People are choosing the way we have to work, rather than the other way around”. The following
158/211 D2.3: Impact of social media, Version V1, PU box provides an example of the way conflicts can arise due to lack of communication between the social media team and dispatchers during an emergency.
159/211 D2.3: Impact of social media, Version V1, PU EXAMPLE: Industrial fire An example of difficulties that can arise between the social media team and other members of the service was provided during a recent industrial fire. Once again information about it was shared on Twitter within minutes of it starting: “There was a big industrial fire and the dispatching centre were upset because our communications people started tweeting about it before they’d had any calls about it – e.g. ‘1km away from the fire, close your windows’ – so people started calling into the dispatch centre from all around the city, asking: ‘We just saw your tweet. Do we need to close our windows too?’ But they didn’t have this information yet, we were ahead of the game and so they were upset about it. But I told them: ‘This just shows that you need to monitor social media’. But it also means we need to communicate better with them”. Social media users and uses Which types of staff or other stakeholders are most likely to use social media during an emergency? As outlined above, the AFS has permanent staff with key responsibility for gathering and sharing information on social media. Any relevant information is shared with others in the organisation to provide situational awareness. They have also recently started an experiment to look at new ways of communicating with journalists via social media during emergencies – so they are given access to a WhatsApp group and they can communicate with the AFS via that group. They also have agreements with newspapers so they use AFS tweets in their communications: “They embed the tweets in the articles – they push our communications through their channels”. What types of information are emergency services most interested in? Overall, social media is used on a day-to-day basis to collect two types of information: a) Big Data: to detect trends and to inform the big picture about an event b) Micro-information: ‘to find the needle in the hay stack’ – i.e. to find individual bits of information, such as videos or photos or descriptions, that can aid the emergency response. More specifically, during a recent storm social media was used to explore the impact of the storm/damage reports (What damage had it done? Where? In what form?), situation assessment (Are there any videos or photos to show that is happening?), and to ask for and share information. So, for example, they saw that people were sharing a picture of trying to put a tarpaulin on a damaged roof to protect a historical building. “So, we contacted them via social media and said: ‘Wait for us, we’ll do it. It is dangerous – please direct message us’. Then we swapped numbers and spoke to them directly to arrange a resolution”. They also used information from social media to find out the impact of the storm – this included for example images that showed that the storm had caused trees to lose their leaves: “We saw pictures of leaves everywhere – we saw that they would block the drains. We normally have trucks to clean the sewers but it was a national holiday so police were sent to get the truck drivers – the clear image of this problem came from social media – this is something you wouldn’t think of but if you see the pictures you do”. After the storm they also asked for specific information via social media to assess the impact of the storm: “We asked for pictures of the hail to help us make a picture of damage in the city”. What kinds of social media citizen roles would be most useful for emergency services? The AFS currently does not use citizens to analyse social media data. However, they do have links with established citizen groups to use their support in emergencies (but not related to social media): “We have contact with an established movement – Antwerp Speaks – a conglomerate of all citizen organisations. We have good contacts with them and we do things about flooding with them”. This includes, for example, during a flood to ask neighbours with pumps to help out elderly citizens who have water in their basement.
160/211 D2.3: Impact of social media, Version V1, PU Information validation What procedures and tools are used to validate such information efficiently and effectively? Reliability in the sense of now being able to trust the information provided via social media was not seen as a great issue for the AFS. First, they were not aware of any examples affected the fire service of false information being shared deliberately – as has been done in other emergencies, such as the Brussels bombing where videos of another bombing was reposted online, “but it’s something you recognise immediately if you’re local’ – we could see it wasn’t Brussels airport”. Second, they felt that any photos or videos posted on social media could be triangulated with other information to check their authenticity: “When you have a picture, you can judge it by looking at the weather and the location … so you can see if it has been taken on the day in Antwerp or not … so it’s very easy to determine if it’s real or not”. Information is also taken more seriously if it is corroborated by other sources of information or social media messages: “If you have three pictures of a collapsed bridge, then yes the bridge has collapsed. But if you only have one, it could be photo-shopped” How can reliability and accuracy be supported, technically? The AFS do not use automated data analysis to validate social media information. As outlined above, information validation is done through triangulation and visual checking of information (photos, videos). Emergency service staff skills and resources What skills are required to help emergency service staff to find, interpret and make use of social media data? One challenge is involving staff member in using and analysing social media who are not familiar or comfortable with this medium. Training can only go so far to overcome such a barrier: “You need a special type of person involved – we need people who are digital natives. Training people who are not is very hard”. Moderating citizen communities As noted above, the AFS already works actively with citizen groups to engage them in supporting other citizens during emergencies – by posting messages on social media. All of this depends on the emergency service to have access to a lot of followers on their social media channels, currently via Facebook and Twitter. However, AFS staff were aware that “social media is used by certain demographics – it discriminates against those who don’t use it” and that they need to work hard to get as many followers as possible during normal operations “to reach a lot people in an emergency”. This is done by sharing engaging information with citizens whenever they can: “We share a lot of behind the scene information – everything we do that is interesting we share – cool photos, cool videos – otherwise we can’t use it in emergency. On Facebook you can only have good reach by having good content”. The ASF is currently considering the use of a bot on Facebook Messenger to answer simple questions during an emergency, so that people can get a quick response to queries. This is not seen as a replacement for 112 calls: “People who have a fire in their homes will not use messenger – they will call. But people with less urgent needs may use it”. The advantage of Twitter is that journalists and other ‘influencers’ use it a lot – so it can be used to share information more widely via these people. But, in terms of reaching the public directly, Twitter is not a good channel: “Twitter has really gone down in terms of its popularity – young people don’t use it, they prefer Snapchat and Instagram – but they are much more difficult to monitor than twitter for our purposes”. But the popularity of Twitter increases during large emergencies, such as the Brussels bombing, because it is an open channel of communication.
161/211 D2.3: Impact of social media, Version V1, PU Key Conclusions The Antwerp Fire Service is a good example of the successful embedding of social media within the day-to-day operations of a fire service. A significant contributing factor to this success has been the fact that a senior member of staff with support from his superior was committed to using social media data in the organisation’s daily operations. They were also able to build on recent examples where information from social media had been used to provide useful situational awareness. The service uses a particular tool, Coosto, in collaboration with other local actors – the police and city council – to detect and share relevant information with each other. However, they are considering adopting a different, more expensive, tool that has been developed specifically for emergency services to overcome some of the issues identified with Coosto. Most senior staff in the organisation are committed to the use of social media information in their operations – however, among operational staff there is still some resistance to its use. Twitter, WhatsApp and Facebook are used most extensively to engage with citizens and other stakeholders – for these to be effective channels of communication during emergencies the AFS realizes the need to attract followers during normal operations via the sharing of interesting and engaging content The main barriers to the successful implementation of social media in emergency services are: the cost of social media analysis tools, difficulties in using the existing tool to identify relevant information, some emergency service staff being resistant to the use of information found on social media, and staff who are not ‘digital natives’ and a trend in using new types of social media that cannot be monitored so easily (such as Instagram and Snapchat) Themes and Questions for Case Studies Round 4 What can the EmerGent tool do that Coosto, PublicSonar cannot do and what is its cost in relation to these other services? Is the EmerGent tool missing any functionality provided by such other tools? How can emergency service staff with negative attitudes towards social media be convinced of its usefulness? What implications does the decline of Twitter usage and the growth of other, less open, social media applications (Instagram, Snapchat) have on monitoring social media during emergencies?
162/211 D2.3: Impact of social media, Version V1, PU Appendix 5: The National Crisis Center of the Federal Public Service, Belgium (Case Study Report) EmerGent Case Studies: Round 3 Background and Context The Belgium National Crisis Center of the Federal Public Service (FPS) Internal Affairs provides its services to the Federal Government and local municipalities in dealing with and responding to crises. One of its main tasks is to collect, analyse and disseminate the necessary information to the political and executive authorities. The Crisis Center also offers its infrastructure and know-how in interdepartmental management and crisis coordination at the national level. During a crisis, the Crisis Center will provide advice on what, when and how to communicate information to the public and discuss the overall communication strategy with the crisis team. The communication service of the Crisis Center employs seven staff full-time (up from four staff in March 2016) and can also draw on the support of a wider team of experts (Team D5) based in local municipalities – who can provide support for crisis communication. The Crisis Center can also draw on the services of a contact call-center which citizens can call to share and receive information during a crisis. On 22nd March 2016, the Crisis Center was hosting a planned crisis communication exercise with the support Team D5 – which meant that 25 members of the Team D5 were on the way to the Brussels head-office at 8am, just as the attacks at the airport started on that day. The BBC describes the events on that day in the following way1: “Two explosions, moments apart, tore through the check-in area of Zaventem airport at 07:58 local time (06:58 GMT). One eyewitness reported hearing shouts in Arabic just before the explosions. Officials said the bombs were detonated just seconds apart at opposite ends of the departures hall. Witnesses said people ran from the site of the first blast, only to be caught in the second, near the main entrance.
1 http://www.bbc.co.uk/news/world-europe-35869985
163/211 D2.3: Impact of social media, Version V1, PU
Just over an hour later, a further blast happened at the Maelbeek metro station in the city centre, close to several EU institutions. A three-carriage train was just leaving the station in the direction of Arts-Loi, the next stop a short distance away, when the blast happened, according to Brussels transport operator STIB. The bomb was apparently detonated in the middle carriage, which was running along the platform at the time. The driver immediately halted the train and evacuated the carriages”.
Social media tools, platforms and technologies What types of social media tools, platforms or technologies are being used by emergency services? During the attack, the 25 personnel in the Crisis Center looked at a variety of media, including the national and local press and social media, as well as listening to the radio/watching TV. However, “in the first hour social media is one of the most important channels for the crisis communication strategy” due to its immediacy. The information from these different channels was collected together using Trello1 - as it allowed them to work together on one document simultaneously. The crisis communication team of the Crisis Center uses a variety of free and paid for tools to collect and analyse social media (and other) data during a crisis. This includes free services such as Tweetdeck and Hootsuite, as well as Coosto (which they have to pay for). These services allowed them to collect a lot of social media messages during the attack – even though ‘It’s not possible to pick up every message’ and also only for particular types of social media. Thus, Twitter for example is used by particular types of people – journalists and famous people – but young people prefer to use WhatsApp or Facebook. However, messages shared on these platforms are much harder to collect and analyse as they are not open access. During the Brussels bombing there were more than 2 million messages on Twitter – and even though it was of course impossible to analyse all of these, they provided useful situational awareness during the event. So, for example, through looking at what people were tweeting about, it was possible to pick up trends in what people were concerned about at that particular moment: ‘What are at this moment, the key concerns of citizens? What are the main questions right now that are bothering them? What is trending on social media?’ This is done by collecting and analysing social media data as well as information from calls to the contact center (in coordination with Brussel Airport Company). For example, at one particular moment when the transport network was shut down there were many messages sent on social media about ‘mobility’ – how to get home or where they were going. Or at about 12 o’clock, four hours after the initial blast, a lot of people were posting messages on Twitter and Facebook about their luggage and cars left at the airport: ‘This was growing and growing. So we knew that we have to make a decision about people luggage and cars and communicate the message with empathy and concern’. Another example is that during the attacks there was a decision initially to close the transport system – this meant a lot of people were stuck in the center of Brussels. It was then reopened but people could only enter the metro or train station by passing through security. This meant that: ‘there were 1000s of people waiting outside railway stations – as we could see from pictures posted on social media. This was a potential risk situation”. The crisis communication team of the Crisis Center was able to identify this issue and share it with the crisis staff by showing them social media messages and photos. Social media is used in a crisis to gather information about what people are saying, but also to send out messages in response to their concerns, and at a later stage to directly communicate with citizens if
1 https://trello.com/
164/211 D2.3: Impact of social media, Version V1, PU they send in questions. ‘An important piece of learning for us has been that social media is not only about sending out information to citizen but also to converse with citizens – people now really expect that. If you send a message to any large company these days via social media, you expect to get a message back quickly. Citizens expect the same in a crisis from authorities’. On the day of the attacks, a team of six people was responsible for the web-care – maintaining the information posted and sent out via Twitter, Facebook and the website. They sent out 220 messages on Twitter and ended up with 50,000 new followers on Twitter and 8,000 on Facebook. The day of the attack the Crisis Center got a call from Twitter asking them what Twitter could do to support them. So they asked them to ‘push’ a message to all Twitter users currently in Belgium telling them to look at the Crisis Center website for official information about the attacks. Google also provided a link to their website below the search box, and Facebook activated the Facebook Safety Application.
How is it used on a day-to-day basis? Coosto is not used on a daily basis – it is only used if and when there is a crisis. This was seen as a disadvantage, as staff were not sufficiently experienced in using them. This is because the Crisis Center is only activated if there is ‘a huge crisis – if, for example, there is a fire in a port, we don’t get involved. Others can handle that on their own’. On the other hand: ‘If a local authority can’t cope the crisis communication on their own, they can ask for support from Team D5, but the Crisis Center itself will only get involved if it is a major crisis such as a terrorist attack or nuclear incident’. What is seen as the strengths and weaknesses? Coosto was seen as a useful tool: ‘It’s a good tool – it’s OK – it gives you a clear overview of the most used words or phrases. For example, ‘mobility’ or ‘luggage’ – which was useful for us to identify the main concerns’. The other free tools were useful as well although they were less sophisticated than Coosto. With Coosto the main challenge is to find a good search string to make sure you get the most relevant information. They also did this by getting it to look for messages using the most commonly used hashtags – although these changed as initially they focussed on the airport and then changed as a result of the metro attacks.
165/211 D2.3: Impact of social media, Version V1, PU Organisational structures and facilitators How is the use and analysis of social media integrated into the emergency service’s organisational structures and systems? It is fully integrated in the working of the crisis communication team– in fact, it is one of the most important ways they look for information in a crisis, as it seen as a way of enabling them ‘to pick up the concerns of the citizens’. What are the organisational resistance dynamics and barriers to the use of social media and how were these addressed? There are no obvious ones in this case. Social media users and uses Which types of staff or other stakeholders are most likely to use social media during an emergency? As outlined above, the Crisis Center has permanent staff with key responsibility for gathering and sharing information on social media during a crisis. What types of information is the crisis communication team most interested in? The crisis communication team looks at social media during a crisis for three types of data: Information: How many victims are there, where is it happening, etc.? Behaviour: What is the population doing right now? Sense-making: What is the emotional impact of the crisis on them? What are they most worried about? So the Crisis Center looked for information and also communicated information on the three levels – information, behaviour and sense-making, using the motto: ‘We know, we do, we care’. What kinds of social media citizen roles would be most useful for emergency services? The Crisis Center did not specifically ask for information from citizens but draws on information shared between citizens on social media as well as any concerns directed at them via social media or the contact centre. Once again, they are most interested in things that they can use according to the three categories above – information, behaviour, sense making. Information validation What procedures and tools are used to validate such information efficiently and effectively? Given the main focus on sense making during the attacks – what were citizens’ concerns – this was not really seen as a major issue: ‘Of course there were a lot of false rumours, but this was not really an issue’. In another recent crisis, it was a bit of an issue – after the Paris bombing, there was a lockdown of the transport network in Brussels and a lot of rumours circulated on social media which were also reported on television and the radio, about police and military interventions going on. In order to cause no harm to the ongoing police operations, people started tweeting cats which has been perceived as a supportive action and sense of solidarity: ‘So one minute, there were lots of messages about police interventions and then the next minute, all you get is cats. And if someone posted a message about a police operation, others quickly criticised them and told them to stop’. So it was kind of a self-corrective mechanism (see article Self-Correcting Mechanisms and Echo-Effects in Social Media: An Analysis of the ‘Gunman in the Newsroom’: http://www.google.be/url?sa=t&rct=j&q=&esrc=s&source=web&cd=3&ved=0ahUKEwjB5ePwiorRAhVJc FAKHYKDBkMQFggrMAI&url=http%3A%2F%2Fwww.logeion.nl%2Fl%2Fmailing2%2Flink%2F5149eda9-
166/211 D2.3: Impact of social media, Version V1, PU fa79-4a5f-85d4- 8aa489350105%2F13555&usg=AFQjCNFOCI7GfWICwfKqVaWNiOQoGmyOpw&sig2=rkCYUsN6q4HbM- l6ro8h8A)
Emergency service staff skills and resources What skills are required to help emergency service staff to find, interpret and make use of social media data? All staff employed in the crisis communication team of the Crisis Center and the members of Team D5 are knowledgeable when it comes to social media: ‘Everyone speaks the ‘same language’ – we use the same methodology, the same tools’, based on training received and regular exercises. Moderating citizen communities The Crisis Center does not have links with citizen groups to engage them in collecting or analysing social media data during a crisis. At the same time, they are very conscious of the help provided by citizens during crises and will tweet about this: ‘We’re very thankful of actions of citizens and we have sent out messages about this’. Key Conclusions The Crisis Center provides a good case study of the way social media is used both to collect and share information during a crisis – as such it is fully integrated in the operations of the center during any major incidents. The center uses both free and costed tools to collect and analyse the data – this provides useful information to pick up the main concerns of citizens during the crisis; at the same time, it was recognised that these tools only capture information from certain types of social media that are easily accessible. This means that information shared via other types of social media more commonly used by particular demographics (such as WhatsApp by young people) are not picked up All staff in the organisation are committed to the use of social media information in their operations. Social media data is seen as providing useful data relating to three categories: information, behaviour and sense-making. These are all used to inform the crisis team and to communicate relevant information to citizens.
167/211 D2.3: Impact of social media, Version V1, PU Appendix 6: Greater Manchester Police (Case Study Report) EmerGent Case Studies: Round 3 Background and Context
Greater Manchester is a metropolitan county in North West England, with a population of 2.56 million (the third most populous conurbation in the UK). It encompasses one of the largest metropolitan areas in the United Kingdom and comprises ten metropolitan boroughs.
Greater Manchester Police (GMP) is the second largest Police Force in England and Wales and has around 6,000 police officers and 3,000 police staff. This is a reduction of 2,000 staff from a few years ago. GMP is organised into 10 policing territories, covering the 10 local authority areas. Overlapping these area Forces, GMP also has nineteen Branches responsible for Force Strategies and the delivery of services to front line policing.1
Whilst social media had been used prior to 2010, before then GMP did not have a social media strategy, but the intention was to use social media to make GMP more accessible. Social media was adopted in the context of a national conversation about police related issues which GMP wanted to be a part of. The newly appointed Web Manager did 6 months of research before implementing the strategy. Social media tools, platforms and technologies What types of social media tools, platforms or technologies are being used by emergency services? Since 2010, the primary social media tool used by GMP is Twitter. Twitter has penetrated deeply across the organisation: with over 60 accounts covering all areas of Greater Manchester. The accounts are largely staffed by Police Officers and staff from local policing teams, with as many as 350 people across GMP active on social media. Apart from these Twitter accounts, GMP have over 50 social media accounts on Facebook, which also covers all of Greater Manchester, along with accounts on Instagram, YouTube, Flickr, AudioBoo, Kik, LinkedIn, Pinterest, Google+, Snapchat, and GoBubble. Many of these platforms are used for specific functions, such as GoBubble being used by local school children during preventative work. GMP have also developed an app called Greater Manchester Police which can be used on iOS or Android devices. With over 40,000 downloads, the app is used to broadcast information, fed directly from the website, and also to report incidents.
To manage these accounts GMP have generally used free tools. The main tools used at the time of writing are Crowdcontrol and Tweetdeck. Tweetdeck is used to look at what people are saying outside of mentions, but they are aware they are “missing out a bit” in only using this tool. In common with
1 https://www.bluelight.gov.uk/portal/nwpolice/portal.nsf/vLiveDocs/PD-DEVA-82TLAK?OpenDocument
168/211 D2.3: Impact of social media, Version V1, PU many police forces in the UK, the office IT is generally poor, including the browser used (GMP only recently upgraded to Internet Explorer 11) which limited the type of online tools that can be used. The tools being used to centrally manage these accounts are being reviewed at the time of writing.
Whilst not covered in the interviews, the GMP Intelligence teams are using social media independently of the Digital Services and communications departments.1 Whilst there were no specifics of which tools were used, the Intelligence team do data trawls. Until recently, they had a freeware data mine tool which allowed them to look at big data. How is it used on a day-to-day basis? During high visibility crimes and emergencies, social media is used by GMP to improve awareness and responsiveness to crimes happening in the city. The triggers for GMP’s involvement in crime response can now happen unofficially: citizens will tag the main GMP account about crimes when local forces are unresponsive. Whilst not all tags are relevant, these mentions can help the police become aware of crimes before they get reported and can lead to rapid charges on high-profile crimes.
One high profile example of this was during summer 2016. There was an incident of racial abuse on public transport and, before any report had been made to the police, police Twitter accounts were ‘mentioned’ in tweets. According to a press release by the Police: “The incident happened at around 7.40am this morning, Tuesday 28 June, on a tram travelling towards the city centre and police were made aware after a video of the incident surfaced online. Two men aged 20 and 18 and a 16-year-old boy have been arrested on suspicion of affray. They remain in police custody for questioning.”2 Whilst these tags are not official reports, in incidents which are first reported online, the GMP’s attitude is “We need to deal with it and be seen to deal with it.” In this case, social media helped operationally as the suspects were arrested within 24 hours with the help of social media coverage and responses from the public via social media.
In day-to-day operations, the social media team has very close contact with other parts of the organisation. For the racial abuse case, the head of Corporate Communications called the chief inspector for the area after seeing the video and he came over to look at the video. After viewing the video, the Chief Inspector took action and then reported back to the head of Corporate Communications that the incident had to be dealt with. In other cases, on an event by event basis, press officers are invited to the control room to provide information and assist intelligence gathering.
A large proportion of the social media work at GMP is preventative. They conduct many campaigns, related to regular calendar events such as bonfire night, and others which are driven by GMP such as awareness raising of child sexual abuse, and modern day slavery. Whilst some of this is done for Public Relations this can help operations too. Besides campaigns, the social media team often refer complaints to the relevant channel before they officially become complaints so that the digital team “deal with a lot of things just online.”
Training for using social media happens on a weekly basis, where new social media users of the local teams go to the GMP HQ to be trained by the Digital Services team. The training is an induction during which trainees receive access to their social media management tool account. GMP use Crowd Control
1 The Intelligence team have social media too, and are using it for different reasons: Intelligence use it for monitoring and analysis, they trawl and use it for covert intelligence. Intelligence look at particular SM accounts that might be run by criminals. 2 http://www.manchestereveningnews.co.uk/news/greater-manchester-news/two-men-boy-16-arrested- 11538089#ICID=sharebar_Twitter
169/211 D2.3: Impact of social media, Version V1, PU and all users have individual log-ins, an approach which started in 2010. The training covers basic skills but also trains users on contempt of court and copyright issues. Whilst GMP do not conduct daily checks on all social media accounts, all content posted is audited and every post is marked with the identity of the author and when it was published. The most important daily connections for GMP Digital Services are with local police teams, who will ask for advice, training, and access.
Observation of day-to-day incident management The researcher was allowed to observe the daily actions of the social media HQ team at GMP following the interviews. The layout of the office was a large open space with five departments in the same room grouped in clusters: Digital Services, Internal Communications, Press, Public Relations and Awards. Many of the departmental functions overlapped, particularly Digital Services which was involved in proactive campaigns (PR), staff events (Internal Coms), reactions and appeals for criminals (Press) and long-service recognition (Awards). Press officials were also in the control centre so that there was always a link to the HQ office and the control centre. There were frequent communications between Digital Services and other departments whilst the observation took place. During the observation, the Digital Manager showed the researcher the ‘Tweet deck’ for the main @GMPolice account and the Crowdcontrol application. These were used to monitor and post to followers on all the central accounts. It was clear that most tweets sent to @GMPolice were irrelevant or did not require action. Whilst trawling through recent messages sent to the @GMPolice account, one tweet was seen that needed to be addressed immediately: a hate crime had been reported to the police by the victim which had not been dealt with by local police forces for several days. Having seen the message, the Digital Manager immediately went across the room to the Press officials and asked them to follow up the log number which had been provided. The Press quickly confirmed the case, noticed it was not logged as a hate crime and then passed the case details to the relevant police team. Further checks showed that no officer had been allocated to the crime and that the crime was labelled in the records as an assault. The Digital Manager then asked the Twitter user to DM his phone number. The user replied by publicly posting his phone number online, instead of direct message and the Digital Manager replied immediately to ask the user to remove the number from public record to safeguard the victim. Finally, the Digital Manager emailed the relevant local Police Inspector about the incident and asked them to deal with this as a priority. Throughout the response there were whilst constant communications with the Press team.
What are seen as the strengths and weaknesses of these?
Some of the weaknesses related to the decentralisation. Due to the lack of central control and huge number of accounts, the individual force accounts are only reviewed every quarter. Due to this, sometimes accounts will not follow the advice to post regularly. This will only be picked up during quarterly reviews or when a citizen messages GMP to ask why is the account is not posting.
This decentralisation means that the content of posts is not checked regularly. Whilst most posts have no problems, some will have a negative response from the media or citizens. From the beginning, the GMP board have accepted that some negative responses are inevitable, with some chief officers embracing bad responses and see room for improvement as a learning opportunity.
170/211 D2.3: Impact of social media, Version V1, PU This less controlling approach is usually supported by Digital Services who encourage informality. For instance, the Digital Manager stated that “We don’t want police to say ‘male juvenile’ just say ‘boy’ - it’ll save characters!” This relatively light approach can be a weakness as the media view of how formal the police should be on Twitter is often out of step with the public view and GMP have had bad publicity when the tone of a tweet is too light. There were three clear examples of this from the interviews:
1. In the first example, an officer tweeted that ‘North Korean missiles have been seen in Eccles [a town in West Manchester]’ and whilst the tweet got 400 ‘favourites’ the Daily Mail reported negatively on it. 2. In the second example, one officer tweeted lots of song lyrics from one of the official Police accounts. The Manchester Evening News contacted the GMP Press team to say the officer has been doing this and they would run a small story about it. The officer was relieved of their tweeting duties, but as the officer in question was doing a tweet chat at the time of being relieved, he tweeted publicly to say he had been ‘told off’ and was going off Twitter. Because of the public reprimand, the story got into the Daily Mail and became a big story. 3. In the final case, during a Yanks festival the official traffic Twitter account tweeted a Nazi uniform at the festival. Following this there was a very negative reaction from the Jewish community and GMP had to officially apologise. These issues are seen as part of the social media model used by the GMP: “You’ve got to be prepared that we will have problems and that’s a risk that not every PCC is going to accept.”
Another issue is that some GMP branches want Twitter accounts where there is no operational justification. In all cases the forces and divisions which are on Twitter need to make an operational case which supports the police’s raison d’etre: to protect citizens through having a positive impact on crime and prevention. For example, the Digital Services have blocked the mounted unit from having social media accounts as they have not made a convincing enough business case yet. From the GMP perspective, the case needs to be made so that they could deal with any Freedom of Information (FOI) request of why the mounted division has an account.
Amongst the strengths of the model are cost-effectiveness. The main costs to using social media at GMP are the technology and price of the number of licences but these do not require large investments. The user-led model appeals to both the police users and to citizens who have a more human access point to their local police. Because of this model, Police users are usually very keen to tweet and the amount of users is probably at saturation point: it would be very difficult to expand the breadth at this point. The only ‘missing’ users from the Digital Services point of view are GMP call handlers as there would be benefit in them using social media.
GMPs conversational and transparent approach has been very successful in getting new followers at key moments. A large number of followers are needed who will ‘mention’ the GMP on Twitter when a significant crime occurs as the Force do not have sufficient resources to continuously check social media for developments; their ‘tag and response’ method works well as GMP are so well known on social media.
171/211 D2.3: Impact of social media, Version V1, PU
Twitter day, 2010 Shortly after the Web Manager started in 2010, there was a local and national conversation about citizens not understanding the nature of day to day policing. To help with this, the board and Digital Services decided to have a Twitter day in Oct 2010. During the 24 hour period, @GMPolice tweeted every call that came in to the GMP to show the breadth of issues that the Force deal with. The transparency was an important aspect: “We didn’t want to compromise. My theory on this is that if you do [a tweetathon] to promote your account it won’t work. There needs to be an operational reason to do it. I was very nervous and we were on the BBC breakfast news.” During the day, lots of technical issues made the task more difficult. In total the @GMPolice tweeted 3000 times, so they reached the daily tweet limit and even had to switch computers several times to other accounts and even to off-network computers. Whilst tweeting operational information, it was helpful to have a character limit on the tweets; this made it more difficult to give away confidential information. There were some operational updates that could not really be tweeted (including a shooting), which the media followed up. There were other calls that the public found funny, such as a member of the public reporting a baby hanging over a bridge (actually a dog that didn’t like bridges). GMP did get a call from the Home Secretary during the day asking ‘What are you doing?’ and “we took that call as a sign of success.” The Twitter day was “a massive turning point to us. It had a massive impact. The police chief wanted to get the nature of policing on the agenda and that worked too. We focused very much on impact around communities and preventing crime so it was a justified exercise. We never try to deliberately get followers.” Quantitatively the success of the day was clear: in 24 hours @GMPolice went from 3k followers to 21k followers.
Organisational structures and facilitators How is the use and analysis of social media integrated into the emergency service’s organisational structures and systems? At the time social media was introduced to GMP there was a movement within the organisation to break down some of the hierarchical structures. The Digital Services team unexpectedly managed to be part of that movement: “We didn’t anticipate the cultural change that it brought. It broke down some hierarchies within the police between ranks and divisions.” Events such as the Twitter day required buy- in at a senior level and gave a greater visibility to social media across the organisation. Partly driven by social media usage, GMP renewed their focus on the frontline and the needs of the community that the Police in Greater Manchester serve.
One of the main tools to communicate and interact across the organisation is the Twitter and Facebook analytics which are published every quarter in a full report to the organisation’s board. This report covers all 105 GMP accounts and is not based on number of followers. In the early years, the report writers were encouraged to measure followers but this is seen as being “the metrics culture”. Instead, the quarterly reports measure engagement. The Digital Services team ask
“are they getting contact and engagement in the community. We put out the top performers and explain how they have achieved that. We pick out the bottom five accounts and look at what they’re doing wrong and whether to close the account down. One officer went from
172/211 D2.3: Impact of social media, Version V1, PU bottom rank to top five as he was embarrassed. And those reports go to the bosses and the account users. The way we use it now it’s part of an ongoing conversations.”
Besides the quarterly analytics, GMP also analyse campaigns. Every post related to a campaign is tagged so each project can be evaluated. This includes emergencies: “Usually we use analytics after emergencies; it’s hard to do it during emergencies such as the riots. During the Robert Hart murder investigation we did social media analytics in coordination with the investigation team.” Analytics have different functions depending on the task. For a campaign, the organisation wants to know how far the campaign has gone, what responses they had, and the level of engagement. For an appeal, they want to know the responses and the reach. Analysis of campaigns needs to show the numbers of conversations, so GMP have pulled campaigns forward because of the high level of responses. What are the organisational resistance dynamics and barriers to the use of social media and how were these addressed?
In the central team, the interviewees reported no organisational resistance: after setting up the social media policy, the document went to the chief officer level and had no resistance to its approval. Within 12 months, there was a tipping point where the board fully understood the impact social media could have. Once they realised how could help them to do their job rather than being a novelty, the Digital team got full buy-in from the commanders. The key events were the Tweetathon and the English riots which ‘really helped’ get buy-in. Interviewees state that “We always remember the internal audience, and share the positive news internally. That made a big difference. We saw social media as a useful thing, during the riots it was an information source.”
At local officer level there was some resistance by officers who could not see the value in using social media in their local area. In some of these areas there was a chief officer who would have to convince the officers to use their account. In some of those cases, officers are chosen to take over the Twitter channel despite their disinterest in using social media. In other cases an officer who was responsible for the local social media account will move departments and that account will stop tweeting. In cases when individuals do not buy into social media, the Digital Service can review and make suggestions for improvement or close the account. Overall, the Digital Services have been supported by senior officers and “culturally the organisation accepts the use of social media.”
When GMP introduced the social media strategy they faced some resistance from both within the organisation and the media. One of the strategies used to help with this was to send a reporter – this one from the Manchester Evening News (MEN) – out on patrol with our first tweeting Police Community Support Officer (PCSO) in Didsbury to see how social media would be used on patrol. This was very helpful as they then understood how social media would be used, and that officers would use it as one of the many options they have to communicate with their local communities.
Internally, the success of that first account really helped to demonstrate how social media could be used to support front line activity. A practical example of this was when there had been a student rape in Didsbury and the tweeting PCSO went out into the community for reassurance activity. While out and about he would tweet his location and tell people he was available for a chat if anyone was concerned and wanted some assistance from him. As a consequence “we were then able to start rolling out social media across the rest of the organisation and within a year there was a Twitter account for every neighbourhood team across Greater Manchester.”
“As part of the continued rollout we held workshops for senior leaders within GMP, during which we had some of our tweeting officers demonstrate the ways in which social media had helped them on the front line.”
173/211 D2.3: Impact of social media, Version V1, PU Social media users and uses Which types of staff or other stakeholders are most likely to use social media during an emergency? As stated in the ‘Organisational structures and facilitators’ section, social media analysis and response is conducted in local teams, by the central Digital Services team and by the Intelligence service. These roles stay in place during emergencies, when collecting intelligence (particularly through citizen responses) becomes more important. The Digital Services team sometimes use analytics to assess rumours, for instance when EDL marches happen they run searches and notify relevant teams and the media when necessary information comes up. The Digital Manager always has Tweetdeck open to run searches using multiple communication channels, to assess what’s coming in and going out, monitor hashtags and conduct key word searches. During an emergency he would often be called to the Control Room to feed and check information to operations.
GMP social media during the Manchester Riots (2011) According to the Digital Manager, two of the biggest emergencies during his tenure were the riots of August 2011 the riots and the murder of two female PCs in 2012. The Manchester Riots were prolonged and had an enormous impact on the status of social media internally: “The riots impacted so many people. We used social media extensively during the riots. We were in the command centres and the Chief Officer was with us. Our team had direct access to intelligence, who would tell us would tell us which areas and streets to monitor and we would search and follow what was happening in those areas. We got the CCTV stills of suspects and would put them on Flickr. This was extremely effective and in one case within three mins we got one suspect identified. Then we would feed such information into the intelligence services and that would form part of the operational response. “We could assess and cherry pick what was important; there was a lot of ‘noise’ and, 90% what came up didn’t need responding to. But we responded to a lot of things. From our perspective, it would’ve been a massive mistake to ignore what was happening on social media. We were comfortable about how to use social media, and we decided early on the counteract rumours and reassure people. We could see the impact we were having at that point in real time. We also had press officers in the control room and talking to the media too. We were communicating all the time, and responding as and when we could. We were even saying about what arrests we were making. Just on the day of the riots we went from 21,000 followers to 100,000.”
What types of information are emergency services most interested in? The information needs of GMP depend on the context: campaigns need information on reach and engagement, investigations need citizen responses on suspects, and general queries from the public are useful in addressing complaints from citizens. In exceptional circumstances that require contact with the Intelligence team, live information about a particular area or events is needed. The technology to assess this information is limited by the current budget.
174/211 D2.3: Impact of social media, Version V1, PU What kinds of social media citizen roles would be most useful for emergency services? From the perspective of the interviewees, people who respond to appeals and campaigns are crucial to GMP’s social media model working. Feeding relevant information to the social media teams helps the force to prevent crimes and to ensure the criminal justice system works well.
Besides these community mediators are appreciated as they give us information and also transmit information to more citizens. They also often correct unfair criticism of the Force, “particularly on Facebook, people will criticise us and citizens will respond and correct them.”
Citizen journalists were seen as problematic by the interviewees as they “don’t understand the legal restrictions – for example reporting charges, and then going into contempt of court. Citizens can often over report by naming the victims. We often have to say ‘you’re not allowed to say XYZ.’” Information validation What procedures and tools are used to validate such information efficiently and effectively? Much of the validation is by common sense – for example, if someone is posting information to the GMP and they have zero followers, this is often a sign to the Force that the information should not be trusted. Generally no tools are used and the main problem is information overload. There are currently no criteria for what to reply to, which is partly an issue of validation. How can reliability and accuracy be supported, technically? The GMP have no particular techniques for assessing reliability “beyond looking at the person and their background.” Emergency service staff skills and resources What skills are required to help emergency service staff to find, interpret and make use of social media data? The main skills needed are communication. Due to the amount of information and large numbers of grey areas, having balanced/objective judgement is also seen as important as “you can’t take things at face value.” Technical skills are not a prerequisite as these can be taught during the induction training.
Soft skills are vital at GMP as the philosophy is to be more conversational than directive. The emphasis on soft skills is not uncontroversial and is disputed in other Forces in the UK. For example, for a counter terrorism incident, “the (London) Metropolitan Police gave us some text to share on our Twitter and told us to say “Don’t do this. Do this”. We had to get back to them, saying that we try to be approachable whilst professional and their text was not right. We even try to be chatty if we can. We’re supposed to be approachable, we’re already authoritative as the police, and we don’t need to ram that authority home.” Moderating citizen communities Building community links are ‘a daily activity’ as a police force and this includes the Digital Service. From their perspective, “You can’t start doing social media when the streets are on fire” as the links, credibility and trust will not be strong. Establishing a community presence for the GMP requires daily relationship building. Social Media response is a key part of police responsiveness, particularly because social media users now expect responses. Social media was established at the GMP in part because they have
175/211 D2.3: Impact of social media, Version V1, PU “had issues with in the past with relationships with the public. We’ve had some national research [before 2010] that showed distrust with public services. Whilst there wasn’t a specific reason for this, I think SM is about relationship building and communication. Officers can build relationships with the local communities. Every event has a social media component too.”
Social media has been part of a wider drive to build community links whilst recognising that these links can be helpful and rewarding for both sides. During the case study visit two public ‘hackathons’ had been organised by the GMP to develop apps to identify missing people, one for children, another for adults. The winner of the child’s app development competition had used data to plot where missing people were most likely to be found and where they had been lost. GCHQ were also in attendance and the winning app was seen as being operationally viable by GCHQ. These apps might be used for operations and also helped child and adults test and develop their skills.
Community reporters are also used where possible, and have joined local police teams, tweeting and posting as they ride along. They are then supposed to post an article or blog to give something back. These are difficult to organise because of insurance but community engagement in the digital work is important and often designed into the work: for example, the new GMP website design will request feedback from the social media community before finalisation.
Often though, citizen communities interact naturally with the GMP social media teams. Ignoring these interactions was seen by interviewees as “putting your eggs in one basket. Officers aren’t always out and about so you need citizen input. It saves them time and effort.” Three examples were given immediately for how citizens naturally helped the GMP:
1. One citizen responded to a ‘wanted criminal’ post which led to the capture of the suspect. 2. Another criminal escaped from prison and posted a ‘catch me if you can’ message on social media. Local citizen information given online and offline led to his recapture within 12 hours. 3. Finally, in Bolton there was an arson attack and, after posts on social media by the GMP, the victims received a massive amount of donations. These donations happened organically without GMP direction. Citizen volunteers also come in the form of Community Watch which can be online, particularly on Facebook. These groups are helpful and GMP keeps track of such groups because during events, GMP can link to online community groups in the relevant area at the appropriate stage. Such groups often organise clean ups which can be tapped into by the GMP. Key Conclusions for Emergent GMP are an atypical Police Force in the UK in their embrace of social media and culture of widespread use. There are a few other Forces which have good social media use: for example, West Midlands have officers who individually tweet. Police Scotland were also noted users as were Merseyside police, yet their Chief engage much in social media. The impression of European Police Forces was that in continental Europe there were more restrictions in terms of what can and cannot be done on social media and the casual and loose approach of GMP might not translate well.
For GMP to use the technology produced by Emergent it would need to be free (or very cheap), work on old browsers and needs to have the functionality to post, reply and do direct messages, monitor wider conversations, and upload media images. Of these direct messages are most necessary: emergency services, especially police, need private messaging in order to have confidential conversations. GMP turned to social media in part due to budget cuts as social media has low costs for a potentially high impact and the tool’s price needs to reflect this.
176/211 D2.3: Impact of social media, Version V1, PU Emergent’s general guidance document “could be useful. For ourselves, we have a lot of guidance in place, some will be relevant to forces in the community.” However in order for the guidance to be operationalised, producing the document is not enough: “You need buy-in from the chief forces. You’d need to go through the College of Policing, and they have their own guidance. You’d have to work with them – and The National Police Chiefs council too.”
The Emergent guidance might suggest the GMP’s strategy to be taken up more widely. Their approach of asking the public for information and utilizing this information works well (as it had in Antwerp Fire Department during a recent hailstorm). However, the reasons it works well are that 1) GMP have developed a reputation for been responsive and not arrogant to citizens online, inviting citizen response, and 2) the @GMPolice account has a huge number of followers (almost 400,000 out of a population of 2.8 million) so information can be gathered and shared with little effort needed for outreach. Due to their approachability and scale, during emergencies such as the riots, the GMP were able to effectively direct people away from hot spots, get information from citizens to identify potential criminals, and actively debunk rumours.
Finally, any Emergent guidance should take note of the importance of senior buy in for ensuring social media is well used in an emergency service. From the early stages onwards the Digital Services at GMP gained the buy-in of the GMP Board and the Chief Officer (who was the National Police Chief’s Council Digital lead). This senior engagement made spreading social media across the organisation far easier as tacit resistance was a less viable option for Divisions which were less interested in social media: take up of social media became expected. This senior buy-in was possible through successful negotiation of events such as the tweetathon and the English riots where social media had a clear operational and public relations impact. Also, the strict need for social media accounts to have an operational justification, and the quarterly reports to the board also made senior staff confident that social media was not been used to trivialize the police force.
177/211 D2.3: Impact of social media, Version V1, PU Appendix 7: Frankfurt Fire Department (Case Study Report) EmerGent Case Studies: Round 3 Background and context
Since 1874, Frankfurt am Main (Germany) has a professional fire department, the Feuerwehr Frankfurt am Main, that is one of the biggest and most modern professional fire departments in the world. Employing approximately 1,000 people, this fire department is also one of the biggest public offices in the city administration of Frankfurt am Main (W1). With a population of 732,000, Frankfurt am Main is both the biggest city in the German federal state of Hessen and the fifth-largest city in Germany (W2). Inhabiting around 2.2 million people, the independent town is at the centre of the urban agglomeration Frankfurt-Rhine-Main. Around 2.5 million people (2012) live in the extended city region Frankfurt while the whole metropolitan region Rhine-Main has a population of approximately 5.5 million people. Twelve fire and rescue stations, organized in four groups (Figure 16 1, red marks 1-4), are spread throughout the urban area.
Figure1: Infrastructure of Feuerwehr Frankfurt (W3)
The fire officers are working 24/7 to protect the population of Frankfurt (W1). Besides fire engines and rescue vehicles that are used in almost every operation, several special vehicles and some extra equipment are available at the different locations. Firefighters that obtained a subject-specific training use these special vehicles and equipment to provide efficient and specialized assistance in unusual operations. This includes, for example, high angle rescue, rescue dog units, a technical assistance service, an environmental protection platoon or divers of the fire department. Both the professional fire department and the volunteer fire department, the latter including approximately 900 active
178/211 D2.3: Impact of social media, Version V1, PU members in 28 volunteer fire departments (Figure 1, blue dots) in the different city districts, are the main pillars of fire protection, rescue service and technical danger prevention. In Frankfurt, the fire department coordinates every non-police hazard prevention unit of the German charities that support civil protection:
Arbeiter-Samariter-Bund Deutschland (independent charitable aid agency) Deutsches Rotes Kreuz (German Red Cross) Malteser Frankfurt (Catholic voluntary humanitarian organisation) Johanniter-Unfall-Hilfe (Protestant voluntary humanitarian organisation) Deutsche Lebens-Rettungs-Gesellschaft (German Life Saving Association) Technisches Hilfswerk (Federal Agency for Technical Relief)
Furthermore, a youth and a mini fire department and the Feuerwehr- und Rettungsdiensttrainingscenter (Training Centre for Fire Departments and Rescue Services) are affiliated to the Feuerwehr Frankfurt. Consequently, the central control room of Frankfurt is the biggest control room in the federal state of Hessen (W1). From here all firefighting and rescue operations as well as all hospital transports of the urban area are coordinated. It is stationed in the Brandschutz-, Katastrophenschutz- und Rettungsdienstzentrum (Center for Fire Protection, Civil Protection and Rescue Service). There are twelve equivalent work places whereof seven are staffed during the day and five at night. The other five work places serve in exceptional situations and as a failure reserve. Four monitors (table of operations, operation documentation, geographic information system, office communication) and a touchscreen to control different functions are available at every workplace.
The operation administrators process around 800 to 1000 requests for assistance per day. Unfortunately, more than 50 % of the 112 calls are no emergency calls. Main reasons for these false calls are people dialling a wrong number, children playing with a mobile phone or even jokes. However, the administrators must not only process the requests for assistance via the emergency hotline but also around 2,500 additional administrative phone calls per day. Further, they are responsible for forwarding different concerns to the relevant departments. Annually, there are around 100,000 alerts concerning rescue and ambulance services and approximately 8,000 assistance services and 6,000 fire operations are necessary.
179/211 D2.3: Impact of social media, Version V1, PU Figure 2: Website of Feuerwehr Frankfurt: Operations Section
In terms of online communication, the website (http://www.feuerwehr-frankfurt.de/) serves as a central contact point that contains general news, current and performed operations, job opportunities, educational media files and safety information1 (Figure 2). The Feuerwehr Frankfurt furthermore maintains several social platforms (Facebook, Instagram, Twitter and YouTube) and an RSS feed which are used by the press service for communicating operational and organisational news.
Social media tools, platforms or technologies What types of social media tools, platforms or technologies are being used by emergency services? The opening of the European Central Bank (ECB) in 2015 in Frankfurt am Main was a significant event that made the Feuerwehr Frankfurt realise the need to take the use of social media for communicating with citizens more seriously (I1, 14:05). During that event, multicultural protesters not only attacked police but also relief forces like the fire services and with the help of social media communication they were able to reduce the attacks on relief forces. They now see social media as a central part of its daily routine: “We consider the use of social media to be essential.” Therefore, they differentiate between the real and the virtual situation of an operation. The real situation includes, for example, actions undertaken by emergency personnel in the context of an emergency. The virtual situation depends on context and public interest. To react to the virtual situation, the Feuerwehr Frankfurt is developing different extensive concepts and strategies that compromise:
1 Covering the following topics: Emergency calls, fire protection, smoke detectors, rescue locations in the Main area, maintenance and information (phone service, speakers, radio, Twitter), fire protection education and instruction, emergency-specific information (advent- and Christmas time, flood, New Year’s Eve, power failure, severe weather; each with an educational video)
180/211 D2.3: Impact of social media, Version V1, PU a) The use of social media as an important channel to actively communicate with the public (citizens and other stakeholders) on both an operative and organisational level. b) The monitoring and analysis of public online behaviour in different contexts. c) Lead communication as a central measure that regulates the virtual situation and prevents it from having negative impact on the real situation.
For operational communication purposes, Twitter (https://twitter.com/feuerwehrffm) is used as a central key tool to provide current information about ongoing operations and to warn citizens in dangerous contexts such as fire (K1, Figure 1). In March 2015 and a few days after the ECB riots, its use was promoted with a 24-hour campaign using the hashtag #24h112 where Feuerwehr Frankfurt posted tweets about their daily routine: everyday life, operations and training units (V1, 12:10). The campaign led to an increase of 2000 followers, earned positive feedback by the audience and media attention; in January 2017, the account @feuerwehrffm has a reach of nearly 40.000 followers. Especially press and journalists use the provided information as source material. The use of Twitter in daily work is intended to facilitate its use during an actual emergency or crisis as both the organisation and the media is prepared for using it.
Figure 3. Twitter account of Feuerwehr Frankfurt. However, regarding traditional organisational communication (I1, 03:15), for example in order to recruit talented young employees, the Feuerwehr Frankfurt also uses Facebook (https://www.facebook.com/frankfurt.feuerwehr), Instagram (https://www.instagram.com/feuerwehrffm/) and YouTube (https://www.youtube.com/user/FeuerwehrFrankfurtaM). Especially Facebook was intended to represent Feuerwehr Frankfurt as an employer and in 2015; for instance, Markus Röck reported the receipt of about 800 applications for 40 open positions (V1, 08:00). The recruiting was furthermore supported with paid advertisement campaigns on Facebook. Contrary to the Feuerwehr München (Fire Service Munich), they deliberately decided to not put operation reports to Facebook, because they saw conflict potential between the reports, which may contain information about deaths, and the like button mechanism, and therefore anticipated a high effort of moderation. Instead, Feuerwehr
181/211 D2.3: Impact of social media, Version V1, PU Frankfurt directs all communication to their lead channel Twitter (I2, 09:20). Instagram (Figure 4) and YouTube, then again, depict the different units of the fire department, their areas of responsibility, interviews and practical exercises. The same applies to Facebook, where additional pictures provide insight into different exercises. On an operational level, these tools are not relevant contrary to Twitter. The respective apps are used for the different channels.
Figure 4: Instagram account of Feuerwehr Frankfurt
The Feuerwehr Frankfurt also maintains a website with the goal to inform people and to facilitate communication. For example, people can access different manuals, instructions and brochures. They can even download an emergency fax form for deaf-mutes or they can inform themselves about ambulance services or the fire department and its organisational structure and history. In the category “communication”, the website provides reports about and pictures of several operations and news concerning, for example, new products on the market and developments in the organisation’s environment. Furthermore, it informs its visitors about upcoming events and introduces the different people responsible for everything in relation to press and external communication measures. In Twitter, links to operation reports and, also in Facebook, security advices of the website are often embedded into messages.
Table 1. Social Platforms of Feuerwehr Frankfurt (4th of January, 2017) Facebook Instagram Twitter YouTube Website First activity February October August 2011 August 2012 December 2013 2016 2011 Reach 61808 3238 39979 2169 442482 subscriptions subscriptions followers subscriptions visitors Publication 542 posts 39 pictures 2106 tweets 26 videos 215 news Purposes Public Public Operational Public Public image, image, communication image, image, recruitment recruitment recruitment recruitment, safety information
182/211 D2.3: Impact of social media, Version V1, PU For monitoring and analysis of social media content, Feuerwehr Frankfurt uses TweetDeck1 due to its real-time capability in terms of monitoring keywords and hashtags and free access. However, it is not used 24/7 because emergency calls are assumed as staring points for monitoring in relevant situations. Since the volume of social media data allows no monitoring of unfiltered data, Feuerwehr Frankfurt uses different search strategies to find essential and relevant information with TweetDeck (I2, 03:40):
Reducing the number of sources, e.g., monitoring important media agencies only. Filtering for tweets with a certain propagation, e.g., tweets that were retweeted at least 10-20 times. Using keywords with query operators if the user is familiar with it and the emergency allows spending time on refining the search term. Finally, for internal direct communication, the Feuerwehr Frankfurt uses messengers like Telegram that are encrypted and meet the industry standards in terms of messenger communication (I1, 26:36). In summary, whereas in the past mainly press conferences and telephone calls were used for this purpose, today online communication is the targeted central component of emergency communication, even if press conferences still are part of it. What are seen as the strengths and weaknesses of these? Main strengths of the social media used can be illustrated by the example of social media use during the opening of the ECB in Frankfurt in March 2015 (I1, 14:05). After some protesters attacked emergency personnel of the Feuerwehr Frankfurt, they used the social networks to encourage the people to not attack the emergency personnel again. People in the web, even activists, shared this call on their sites. Consequently, attacks against the emergency personnel stopped and foreign groups were made aware that the fire department, relief forces and police are different departments in Germany, leading to regret of their behaviour. These results highlighted an important strength of the social media use: “These networks provide the possibility to influence and control real-life behaviour.” Thus, the use of social media has a positive retroactive effect. Another positive effect was shown in the context of refugee aid in September 2015 when social media calls were used to mobilize spontaneous volunteers and delegate many suitable tasks accordingly (I1, 14:05).
Table 2. Pros and Cons of Social Media Use Pros Cons Influence and control real-life behaviour Limited personnel and time resources Rapid delivery of and response to Pressure due to fast information response information Reach, e.g., providing information for journalists and press Personnel recruitment Public image maintenance Operational communication (esp. Twitter) Integration of spontaneous volunteers
1 TweetDeck is a social media dashboard application for management of Twitter accounts that makes it easier for users to track the real-time conversations they care about. It brings more flexibility and insight to power users through a customizable layout.
183/211 D2.3: Impact of social media, Version V1, PU Thus, the technological dimension of the communication with citizens is linked with both advantages and disadvantages (Table). On the one hand communication and coordination of the public via tools such as Twitter and Facebook is often much faster than via traditional channels of communication. It supports rapid delivery and response. However, on the other hand this also means that people that are dissatisfied with the work of them or disappointed can also react faster and distribute their opinion on the internet. This is especially negative in cases where danger prevention uses channels that are not fast enough to counter harmful information. Hence it can be said that the increased speed of communication puts the Feuerwehr Frankfurt under pressure. Furthermore, the Feuerwehr Frankfurt does not have sufficient personnel resources and thus has problems to invest the time needed (I1, 20:39).
In terms of the social media tools, Feuerwehr Frankfurt values the real-time capabilities for monitoring keywords and hashtags, integrated filtering options and free access of TweetDeck. They also analysed commercial solutions such as Brandwatch1 which are too expensive and whose target groups are not organisations such as fire departments (I1, 06:50). However, TweetDeck only supports Twitter and citizens may publish relevant information, including media like pictures or videos, on different social media. According to a press spokesperson, the “optimal” social media tool should provide (I2, 06:28):
A good overall usability Capabilities to pre-structure actions and content The flexibility to adapt it to the current emergency Support for all relevant social media Possibilities to capture the mood of citizens He concludes that current systems are too expensive, not operational enough and more suited for medium- or long-term observations but not emergencies. Organisational structure and facilitators How is the use and analysis of social media integrated into the emergency service’s organisational structures and systems? Social Media is a central part of the Feuerwehr Frankfurt’s organisational structure that is integrated in their everyday practice and their operational plans. Three departments are subordinated to the head of office: 1) Press and public relations, 2) staff council and equal opportunities officer, 2) hazard prevention (red: training centre, prevention and planning, operation) and 3) infrastructure (green: logistics, administration, information and communication). The access to and use of organisational social media accounts is restricted to press spokespeople working for the press and public relations department: “Just like management and environmental protection services, press spokespeople are part of our organisation. They are a special unit” (I1, 12:49). The respective staff are trained appropriately to use social media as a standard procedure in emergencies. Due to their on-call duty press spokespeople are available 24/7 and used in the executive staff during emergencies. As press spokespeople are connected technically via laptops, mobile phones and tablets, they can immediately react to emergencies, regardless of whether the person is at home or at the deployment site (I1, 06:50).
1 Brandwatch is a social media monitoring tool, which archives social media data in order to provide companies with information and the means to track specific segments to analyse their brands' online presence. The tool's coverage includes blogs, news sites, forums and social networks such as Twitter and Facebook.
184/211 D2.3: Impact of social media, Version V1, PU
Figure 5: Organisation of Feuerwehr Frankfurt Another factor is that almost everybody within the organisation uses social media privately and thus the whole workforce operates as a “search algorithm” and contributes to monitoring activities (I1, 11:16). They are convinced that at least one of the 1900 people in the organisation finds something of importance. As a further measure the Feuerwehr Frankfurt is planning to build virtual teams that, together with the management, can cope with a virtual situation (I1, 03:15). However, considering this as a plan, the organisational embedment is not jet elaborated. To what extent does this facilitate or hinder the use of information gained from social media by these organisations? The 24/7 availability of press spokesmen allows operate social media, especially Twitter, in every emergency. A further facilitating factor is that, although the fire department is a strongly hierarchical organisation, they established an organisational structure which allows press spokesmen to disseminate tweets quickly without administrative barriers that could hinder the timely dissemination of relevant information (V1, 30:50). A press spokesman highlights that a progressive organisational culture is required to enable a successful integration of social media. The large size of the fire department, including the volunteer fire departments, also constitutes a large base of trusted users which can contribute valuable information via social media even if they are not directly involved in the emergency (I1, 11:16).
However, the limited budget and personnel resources don’t allow for communicate over and monitor social media full-time, because press spokesmen also have to fulfil regular fire department duties and interact with traditional media (V1, 26:00).
185/211 D2.3: Impact of social media, Version V1, PU What ‘resistance to change’ dynamics in relation to the use of social media can be identified and how might these be addressed? A press spokesman indicated that some employees may not like social media or that feel overrun by the technology (I1, 20:39). However, due to the organisation’s hierarchical structure this is not an issue because the employees follow instructions from the top management which demands the use of social media (I1, 33:42). Therefore, the employees comply with the requirements and do not resist. It is worthwhile to mention that the Feuerwehr Frankfurt, on the one hand, is a big fire department with focus on communication and, on the other hand, hired a former Twitter employee to teach in the communication with social media. Which organisational structures or arrangements are likely to facilitate the use of social media in emergencies? As already mentioned, the 24/7 availability of press spokesmen, an organisational culture and guidelines which allow to interact with social media without administrative barriers, and the large base of trusted users facilitate the use of social media in emergencies, although personnel resources are limited. A press spokesman furthermore indicates that they are creating documents in that formalize existing problems, strategies to overcome them, tasks, measures and tools (I1, 35:10). However, they are more like a comprehensive checklist currently, but an integrated concept, which allows implement an effective and quick use of social media, is still missing. Some core questions are:
What is operatively relevant information in social media? How can we aggregate and integrate the information to improve the picture of the emergency and to influence decisions? What are the potentials of social media in (daily) medium and large-scale emergencies? Social media users and uses Which types of staff or other stakeholders are most likely to use social media during an emergency? From the internal point of view, only press spokespeople have access to organisational social media accounts and therefore are able to use social media over representative, official channels. However, employees especially from the voluntary fire departments can contribute using their private social media accounts by actively sending potentially relevant information or passively being monitored respectively recognised as “search algorithm” by press spokespeople (I1, 11:16). Considering external entities, the Feuerwehr Frankfurt explicitly publishes operational information on Twitter to provide reliable information for the general population, journalists and the press, but also increasing the reach with the presence of the channel in traditional media. What types of information are they most interested in: for example, situational awareness data from citizens or data on the public mood or the emergence of rumours or misinformation? According to the current use of social media, Feuerwehr Frankfurt is especially interested the public mood of citizens (I2, 06:28) and potential rumours or misinformation (I1, 03:15) to prevent negative influence from the virtual realm on the real emergency. Furthermore, it is important to identify misinterpretations of the situation to counter a potential loss of trust (I2, 11:26). From the analysis of the November 2015 Paris attacks and 2016 Munich shooting, Feuerwehr Frankfurt sees the necessity to establish a lead communication which means adopting the moderating role in the current emergency and providing information to citizens in sufficient quantity and speed. To establish the lead in such
186/211 D2.3: Impact of social media, Version V1, PU situation, the trust of the population is required, but it is seized as an opportunity to control and counter rumours or misinformation (I1, 14:05). What kinds of social media citizen roles would be most useful for emergency services? As shown in the example of the ECB opening, for the Feuerwehr Frankfurt it is most useful if citizens distribute information shared by the Feuerwehr to persuade people to do the right thing (I1, 14:05). Because of insufficient personnel resources it is not appropriate to use social media „to keep the citizens onboard” and to organize groups of volunteers: “Spontaneous volunteers are employed spontaneously and otherwise they should become part of a volunteer fire department”. Information Validation How can emergency services be persuaded that information provided via social media from citizens is credible and trustworthy? As indicated in section 0, due to the organisation’s hierarchical structure the persuasion of the press spokesman is not an issue because the employees follow guidelines and instructions from the top management which demands the use of social media (I1, 33:42). Therefore, the employees comply with the requirements and do not resist. What procedures and tools are used to validate such information efficiently and effectively? Although TweetDeck is used to monitor and analyse social media, the Feuerwehr Frankfurt does not use a special tool with information validation components. However, a press spokesman was really interested in the information quality component of the EmerGent architecture and its algorithmic backend which is a missing feature in their architecture (K1). In a research project, they furthermore examine an information quality approach for filtering social media content based on the 1) geolocation of the user, 2) credibility of the source, 3) relevance of the source, 4) credibility of information and 5) relevance of information (I2, 15:36). Using this approach, each component is assessed with a score between 1 and 4 and the person responsible for emergency communication can define a threshold to filter information: “At the beginning of an emergency, I just want to see information with a relatively high score in terms of credibility and relevance. With an increasing number of staff involved, however, the threshold could be reduced to get a more comprehensive picture of the incoming information” (I2, 15:36). How can the reliability and accuracy of social information be supported from a technological and technical perspective? Tools must support the fast identification of low quality posts, misinformation and rumours to implement a fast information policy which allows to counter these (I1, 14:05) and to establish a lead communication (section 0). Staff skills and resources What skills and resources do emergency services need to use social information efficiently and effectively? A general rule of Feuerwehr Frankfurt requires that personnel must perform communication tasks in the daily routine to be prepared for communication during emergencies (I2, 28:06). Some of the relevant skills identified by Feuerwehr Frankfurt to find, interpret and use social media are a) an understanding of different social media networks, corresponding tools and their characteristics, b) skills to keep up with the development of technology, its new variations and fields of deployment, and c) socio-psychological strategies on how to counter irrational behaviour in social networks. From the
187/211 D2.3: Impact of social media, Version V1, PU perspective of resources, multiple positions are required that explicitly deal with the topic of social media communication to ensure a good social media performance continuously as well as to develop and improve best practices and guidelines (I1, 20:39). Do they have these skills and resources? As already indicated, a former employee of Twitter was hired by Feuerwehr Frankfurt to impart her knowledge to the relevant employees. Working on a temporary contract, she is now responsible for increasing the employees’ skills. Aside from this, it is very difficult to expand the employee’s expertise due to the rapid development of technology, new variations and fields of deployment over the recent years, making it hard to exploit the maximum potential of social media yet (I1, 20:39): “Tactic without technology is helpless, but technology without tactic is useless.”
Furthermore, the necessary personnel resources are only available on a very limited basis because most of the personnel perform their primary tasks at the fire stations. According to a press spokesman, the inertia and the salary structure of the public service prevents an effective and efficient competence development. Although there is a motivated team dealing with social media, administrating daily work is very time-consuming so that “it is very difficult to invest the time necessary to deal with social media conceptually” (I1, 20:39). Therefore, Feuerwehr Frankfurt also considers the deployment of Virtual Operation Support Teams (VOST, see section 0). What guidance or training would be most useful to help emergency services use social information efficiently and effectively? Due to their communication focus and the expertise from a former Twitter employee, Feuerwehr Frankfurt feels confident in the use of social media from a technological perspective. However, as a press spokesman outlined, they would greatly benefit from socio-psychological research to develop strategies on how to counter the irrational behaviour that sometimes emerges in social networks (I1, 14:05).
Furthermore, guidelines for the use of social media in emergency situations that are valid across organisations would help the organisational and inter-organisational overcoming of emergencies. A press spokesman proposes the establishment of a central institution which “analyses emergencies, reveals problems, develops solution strategies and recommends tactical measures in terms of a catalogue of measures which is available for all federal states, municipalities and districts” (I1, 20:39).
Moreover, the Feuerwehr Frankfurt wants to improve and enhance the operationalization of their approaches regarding the use of social media. For this, standardised guidelines, or catalogues of measures, would be useful to structure the implementation of the different concepts if they present the measures in a suitable layout that “facilitates the actual use in an emergency”. The internal documents (section 0) they currently use are very detailed and thus not helpful for the implementation in emergencies which require fast but good performances. Despite their interest in such a guideline, a press spokesman emphasizes that best would be if they do not need such a document. The main goal is to “anchor the concept in the personnel’s minds” (I1, 35:10). Moderating Communities How important are public/volunteer social media communities in supporting emergency services? Already during the 2013 European floods and even more during the September 2015 refugee aid, Feuerwehr Frankfurt realised the value of spontaneous volunteers in performing a variety of required
188/211 D2.3: Impact of social media, Version V1, PU “non-hazardous” tasks. In the latter case, volunteers performed some special tasks (e.g., interpreting/translation), but in tendency more general tasks that require no specialization (e.g., supervision) (I2, 31:16). A press spokesman emphasizes that “fire departments don’t just exist since yesterday” and there was always an interaction between citizens and fire departments, but social technologies enable volunteers to organize much faster than previously, which constitutes a novelty (I1, 23:50). However, it seems unattainable to moderate these huge amounts of potential volunteers continuously, e.g., due to the demand of personnel and time resources, and volunteers could just join the voluntary fire department if they want to be involved frequently. How do emergency services engage with these communities? The engagement with communities depends on the scale of the emergency, which can be an isolated event or affect a large area. In the latter case, the federalism of Germany constitutes an issue to create comprehensive interaction concepts: Because fire departments are municipal organisations and, in this case, as soon as the emergency exceeds the borders of Frankfurt, the handling of outside areas come under the responsibility of other organisations which may implement other interaction concepts. To overcome this issue, a central institution for developing comprehensive guidelines, as described in section 0, seems suitable (I1, 20:39).
On the one hand, social media communities are more handled as autonomous units. During the September 2015 refugee aid, for example, Feuerwehr Frankfurt established a tentative operations centre together with the federal police, the German Railways and selected contact persons of volunteer communities. On the other hand, the social media use of Feuerwehr Frankfurt is very regulated: While Twitter is used for unidirectional operation communication, which may comprise advice to social media communities, Facebook, Instagram and YouTube are only used for non-operational communication like public image maintenance and recruitment. However, for improved engagement, Feuerwehr Frankfurt considers the deployment of Virtual Operation Support Teams (VOST), and their concept is discussed in the upcoming subsections. What should be done to support social media volunteer communities in emergencies to help them become more efficient and effective? Emergency services may provide structures which improve and support the coordination and integration of volunteer communities. Although not yet implemented, the Feuerwehr Frankfurt considers the deployment of VOST teams using the personnel of volunteer fire departments or other administrative districts: “Because we have volunteer fire departments all over the town with honorary employees that work in different environments of their regular jobs, we just need to search those with basic skills in communication and the respective technology to build up displaced teams” (I1, 26:36). These displaced teams then could be activated as VOST teams and provided with self-coordination tasks via the internet during large-scale emergencies (I2, 31:16). What can emergency services do to make stronger links with such communities before, during and after emergencies? On the one hand, the Feuerwehr Frankfurt thinks of providing apps or creating Facebook groups as enabling technologies for direct interaction with citizen communities during large-scale emergencies, although the plans are not yet concretized (V1, 33:00). Social media, traditional media and press can moreover provide guidance, increase the public perception of volunteer communities and therefore help maintaining their motivation over an extended period (I2, 33:06). On the other hand, the presented VOST approach (section 0) could be applied to create stronger connections and trust with citizens and citizen communities through mutual support in overcoming the emergency (I1, 26:36).
189/211 D2.3: Impact of social media, Version V1, PU Conclusion Key Facts The Feuerwehr Frankfurt uses Twitter as lead channel for operational communication while complete operation reports are available at (and linked via Twitter to) the official website. Facebook, Instagram and YouTube are primarily used for public image maintenance and employee recruitment, but occasionally for the dissemination of safety advice by linking the respective content from their website. Besides the above-mentioned purposes, the possible influence and control of real-life behaviour, the reach (e.g., in combination with journalists, press and traditional media) and the integration of spontaneous volunteers are recognised as strengths of social media. Contrary, the limited personnel and time resources as well as pressure to act due to fast information response were mentioned as weaknesses of social media. TweetDeck is used for social media monitoring not 24/7 but only on demand during emergencies. Commercial solutions like Brandwatch were analysed but are too expensive and, as they are not optimized for emergency services, don’t fulfil the requirements of Feuerwehr Frankfurt. However, the need for an information quality component became apparent and they are already putting research effort into this issue. Social media is an integrated part of the organisational culture and structure of Feuerwehr Frankfurt. Press spokesmen exclusively manage social media communication directly without administrative barriers that could hinder the timely dissemination of relevant information. The basic guidelines on social media use, the coaching of a former Twitter employee and the hierarchical structure ensure the adaption of social media. The Feuerwehr Frankfurt proposes the establishment of a central institution which “analyses emergencies, reveals problems, develops solution strategies and recommends tactical measures in terms of a catalogue of measures which is available for all federal states, municipalities and districts”. These could help overcome inter-organisational work, especially if an emergency overlaps different areas of responsibility, and improve the integration of spontaneous volunteers. Spontaneous volunteers and social media communities are valued as important stakeholders in overcoming emergencies or large-scale disasters, but due to limited resources it is impossible to moderate them all. However, the Feuerwehr Frankfurt thinks about the implementation of VOST teams based especially on the (personal) networks of honorary employees of the volunteer fire departments.
190/211 D2.3: Impact of social media, Version V1, PU Appendix 8: Graz Emergency Services (Case Study Report) EmerGent Case Studies: Round 3 Background and Context Graz is located in the state of Styria in Austria. With a total area of 127.58km² and a population of 315.464 in 2016 it is the second-largest city after Vienna1. The city itself is divided into 17 districts. The first district is called “Innere Stadt” and is part of the Old Town, which was declared a UNESCO World Heritage Site in 1999. It was mostly in this part of the city that an incident involving a car driver running took place. It started in the 5th district, called Gries. In Figure 2 the numbers 1 to 3 are in Gries while 4 to 9 are located in Innere Stadt.
As the incident took place at 12.25a.m on the 20th of June 2015, which was a Saturday, the streets of the city centre were pretty crowded. In addition to a normal Saturday there were some events related to the 2015 Austrian Grand Prix.
In Styria the system is divided in three parts as shown in Figure 1: The fire brigade, the ambulance service and the civil protection agency. Each of them has their own emergency number and their own control centre. The fire brigade in Styria consists of 692 volunteer fire brigades, 83 volunteer plant fire departments and one fire service in Graz. The ambulance service is called ÖRK (Österreichisches Rotes Kreuz / Austrian Red Cross). In case of a big crisis there is another authority called special department for civil protection and national defence (Fachabteilung Katastrophenschutz und Landesverteidigung). If this authority needs to be activated it can also use the resources of the Austrian militia (Bundesheer).
Figure 1: The structure of the ES in Styria in Austria
In case of a crisis the command structure depends on the extent of the incident. The highest rank in a city is the mayor of the city. Therefor he will have the power of order. If more than one city is affected the district commissioner takes over. If more than one district is affected by the incident, the next step in the hierarchy gets involved. This is the state of Styria with the special department for civil protection and national defence, as mentioned above.
It is important to know that on the 20th June 2015 an event for the high-level firefighters was held. Therefor they were all gathered in the same place when they heard about the incident. As the mayor of Graz was involved in the incident himself, he had made an emergency call and called his college in the
1 http://www.graz.at/cms/beitrag/10034466/605976
191/211 D2.3: Impact of social media, Version V1, PU municipal administration. In the beginning the situation was quite unclear. The people at the event did not know exactly what had happened, they just knew something big happened. The keywords reported were “car”, “shooting”, “knife attack” – which does not seem to fit together. It was said that there were a lot of injured and dead people. So the firefighters and politicians who were from Graz left the event and drove home. The incident On the 20th June 2015 at 12.25a.m. a 26 year old Austrian ran amok with his car. The driver attacked pedestrians and cyclists with his SUV (Daewoo Rexton) with an estimated speed of up to 100km/h (62mph). He killed 3 people (one infant, a 25year old woman and a 28year old man). 34 people were severely injured. The driver did not only use his car to attack people: At one point he got out of the car to attack two pedestrians with a knife. Figure 18 shows the route the driver took as well as all places where he injured or killed people.
Figure 2: The way of the amok driver
7 8
9
6
1 2 4 5 3
No. action place 1 Starting point 2 Car hits 3 people – one dies Zweiglgasse-Lagergasse
3 Car hits 3 more people Augartenbrücke 4 Driver stops, leaves the car to attack 2 Grazbachgasse people with a knife 5 Car hits 2 cyclists Schönaugasse 6 Car hits 1 cyclist Dietrichsteinplatz & Schlögelgasse 7 Car hits several people – two die Hauptplatz Car crashes info a cafes outside terrace* – 8 people injured 8 Car hits 2 more cyclists Albrechtgasse 9 Driver surrenders to the police Police station in Schmiedgasse * at the café there was an event related to the 2015 Austrian Grand Prix held
192/211 D2.3: Impact of social media, Version V1, PU Social Media tools, platforms and technologies At 2.30p.m., 2 hours after the incident was reported, the first press conference was held. During this the managing director of the municipal administration realised, that the use of social media was important. In the chain of command he is the one to decide how and with what content the communication should be done. Therefore he deployed 2 employees of the city of Graz to work on SM and the city’s website. Fortunately, one staff member could be activated who had studied journalism and public relations. So she had knowledge about SM and how to deal with it. Their task was to follow the discussion on social media and to give information. As there were only 2 employees working on SM, it was not possible to take part in any SM discussion. They could only lead the coordination of the communication with the news services.
Due to the shortage of employees for this task, it was decided to use the website “Graz.at” to give news updates to the population as well as organisations. This was the only platform that was used actively to broadcast information. The only addition to the website was a service hotline that had been created. As all the facts and news were presented at “Graz.at” other organisations always hyperlinked their info with the website. In this way “Graz.at” became the centre of the Austrian media landscape during the crises. In that point in time the use of the cities website was the only way to go. Neither the fire brigade or the ambulance service nor the police had an SM presence that could have had the same impact.
Furthermore it was decided to do this from a victim’s point of view, because the city itself was kind of a victim of the violence. This was a spontaneous decision by the one staff member who had studied journalism and public relations. At 5p.m. an online condolences book was released on the website: http://www.graz.at/cms/beitrag/10251769/6497524
As the driver (Alen Rizvan Rizvanovic) was very active on Twitter, the question arose whether the incident could have been avoided. He had 2.578 followers on Twitter. Mostly from Arab countries and one suspected neo-Nazi. One newspaper reported that he had shifted to a stricter interpretation of Islam. He was married with two children. But due to domestic violence he got a restraining order on 28th May 2015. Therefore his weapons licence and semiautomatic firearm with ammunition was revoked by police. He himself said he felt persecuted and this made him run people over. What types of social media tools, platforms or technologies are being used by emergency services? At the time of the incident the ES did not use SM in their crisis management strategy. If they use a certain SM channel they usually give some general information in non-crisis situations.
In his master thesis Thomas Meier researched the use of SM in Styria. In general he observed that SM is mainly used to give information rather than to gather some. The 8 different authorities he looked at use the various SM channels to inform the people in a better way. The authors findings are shown in
Table.
KLS is the department of communication for the state of Styria (Kommunikationsabteilung des Landes Steiermark) LFV is the national fire brigade federation (Landesfeuerwehrverband) ÖRK-LV is the national association of the Austrian red cross(Österreichisches Rotes Kreuz – Landesverband) ÖRK-RL is the headquarters of the Austrian red cross(Rettungsleitstelle) MKS is the part of the militia located in Styria that will provide support in a crisis situation (Militärkommando Steiermark)
193/211 D2.3: Impact of social media, Version V1, PU LPD is the state police headquarters (Landespolizeidirektion).
Table 3: used SM channels in Styria [Meie16] KLS LFV ÖRK-LV ÖRK-RL LPD MKS Facebook x x x x x x Flickr x Google+ x x Homepage x x x x x x Instagram x x x x Twitter x x x x x RSS-Feed x x x x x x WhatsApp YouTube x x x x x Twitter is not as big in Austria as it is in for example Germany. And in Germany Facebook is even more commonly used than Twitter. Hence the presence of the ES on Twitter in Austria is low. What are seen as the strengths and weaknesses of these?
General strengths and weaknesses
At the time of the incident the ES did not use SM in their crisis management strategy. The incident in Graz did make them realize that SM is useful. There are certain strengths that the ES realised:
People are used to SM. Therefore SM is the first place they look for information. If the ES is not using this channel, people will rely on other (maybe less reliable) sources. People tend to “believe in the first message” [Haha14]. Therefor the ES should be the one providing these “first messages” about every new development. People share their feelings and worries about everything on SM. Therefor ES can extract frequently asked questions and be able to answer to them – again giving the first message. Also some general challenge were identified and some concerning C2A resp. A2C communication. These are the general challenges:
If every ES used their own SM channel, there will be the danger of broadcasting information that contradict each other. Maybe the police would give different info than the fire department. If this happens, people will stop believing the posts of the ES. So it will be necessary to coordinate the SM correspondence. SM would not work in a blackout situation The first challenge was not a problem during the Graz incident, because every ES only hyperlinked to the city website “Graz.at”. This was the only place, where new information was given by the authorities.
When it comes to C2A resp. A2C communication the interviewees identified different challenges which are summarised as follows:
C2A communication
One of the main problems is the sheer overwhelming mass of information. If you search for a hashtag or keyword you will get thousands of Tweets, comments on Facebook etc. It would take a lot of time, if
194/211 D2.3: Impact of social media, Version V1, PU somebody wanted to read all of them. So, a system that aggregates the info would be most useful. It would also be interesting to have the possibility to answer to people’s messages. Summarised, the weakness is that you need to know what you’re looking for. Having the right combination of search terms (hash-tags) in time available is a tough job.
A2C communication
If 300 people wrote something about a car in the city centre of Graz it should be possible to write a message to all these 300 people
The feeling in Graz was that it is very important to have only one channel on which people can look for information. All SM accounts used during the incident had always linked to graz.at. The idea was to have a place to collect consistent information.
With this approach, they avoided having to have discussions on different social networks, as they would not be able to get into the discussion because of the small number of employees. The fact is that only 2 employees were working in this field. Organisational structures and facilitators How is the use and analysis of social media integrated into the emergency service’s organisational structures and systems?
After the incident in Graz the awareness of the importance of SM has risen. People realised how critical the factor time in communication really is. But they still don´t know how to use SM for their purpose. Therefore SM is so far not integrated in a fixed process: Neither is there an analysis of posts nor a use of SM for communication – neither for critical incidents nor for daily business.
The only use of SM so far is to give information whenever the ES feel the need to communicate. As the impression of the interviewees was, that the media will publish whatever they think (if they do not get information in time), they want to be able to give information in the future to avoid that the press gives false news. Now a SM plan and –strategy is currently in the procedure of being established. What are the organisational resistance dynamics and barriers to the use of social media and how were these addressed?
As in Graz there is no use of SM in emergency situations so far, there are no procedures to deal with the organisational resistance dynamics and barriers. These barriers are not even completely revealed by now. But the interviewees already expressed different worries:
The public usually has a tendency to “believe in the first message”[HaHa14]. So for ES there is a trade-off between time and credibility. If the ES acts to slow they are afraid of losing the sovereignty of information in critical situations and therefore fear to lose control over the situation. This is a strong argument for the use of SM – but there is so far little knowledge about what SM channels could be used and should be used in what kind of situation.
To monitor SM further resources are necessary. “It is unclear who will pay for that”, said one interviewee. The need of money is always an impediment for implementing new things. As SM is not established yet, there is simply no budget for SM staff. The impression was that it might be hard to convince the authorities in charge about the importance of this topic. There is a lawsuit concerning a professional photographer, who takes pictures of incidents for a living. He is strictly against the posting of pictures from ES. If they give the pictures to the media for free, he cannot sell them. This kind of lawsuit gives the
195/211 D2.3: Impact of social media, Version V1, PU interviewees the impression that the authorities in charge will think, that SM brings more trouble than good.
One more concern of the interviewees was the negative impact a post might have: If citizens broadcast a video of an emergency situation there can be a discussion about the tactic etc. of the ES. This can cause a discussion about the “how?” and “who did what and when?” of an operation. In the worst-case scenario, this can end up in a so called “shit storm” against the ES. This possibility makes SM also kind of problematic for an ES and therefor there is a suspicion about whether SM can be seen as a support rather than a problem.
One of the interviewees said, that “someone may think that Austria is all in all always a little bit behind the latest trends. i.e. ES are still organised very traditional.” For this reasons, it is still a process to get the people and authorities to want to implement SM in ES.
In Austria there is also a technical barrier: The broadband connection to the internet is still not possible everywhere. If people are not well connected to the internet, their use of SM will not be as high as in areas where the connection is better. In the not-so-well connected areas people will not be reached by SM posts of the ES. This has to be taken into account if an SM-tool would be implemented. The people living there might get the feeling of being cast out. They could fear, that if the ES uses SM important information will not reach them anymore.
One more barrier is the fact, that “the net has its own language”. There are a lot of new terms and proverbs and ways to express things. If you want to get the right information you have to know how people will talk about it and where they will talk about it. This is also the case if you want to give information. You have to know which words to use to make the people understand exactly what you want and especially if you want them to react in a certain way or want certain parts of the population to know about things, you have to know where to post which info with which words. As SM is not that common in the traditionally organized ES in Austria it would be absolutely necessary to give the ES some tutoring. This might also cause a budget discussion.
In Styria there is a certain way of organizing incidents: In general the ES with the closest relation to the incident should be the one to handle the communication. So if there is a fire, the fire brigade should handle it; if there is a riot, the police should etc. This situation makes it hard to control the information. Therefor the authority is worried about losing control and influence. Especially in parts where the authority has to have the sovereignty of communication (giving an evacuation order e.g.)
All in all it can be summarised, that there are three major issues:
The financing is very unclear The use of SM is not that common in the traditionally organised ES Some parts of Austria do not have a good internet connection Social media users and uses Which types of staff or other stakeholders are most likely to use social media during an emergency?
All in all the interviewees got the impression, that “right now everybody has realised the importance of SM to some extent”. But the affinity for SM of one person itself is the key. If they are into SM, they want to use it in their work in the ES too. Otherwise it is hard for them to really understand its importance and how it works and can be used.
196/211 D2.3: Impact of social media, Version V1, PU As mentioned above it is common that the ES with the closest relation to the incident should be the one to handle the communication. Therefore there is a need for SM staff in every ES – so there are a lot of stakeholders. In addition to that the authority (i.e.. the major of the city) needs to have the sovereignty of communication. So the different stakeholders do have to have a good communication between each other.
In the Graz case the only person they had on their staff for the work with SM, was the one who had studied journalism and public relations. Because of her study her affinity to SM was high and she knew a lot about it. She was able to decide on which channel to communicate and how to communicate. What types of information are emergency services most interested in?
Data and multimedia files which enables ES to improve the operational picture. What kinds of social media citizen roles would be most useful for emergency services?
As the interviewees said, that they would like to get information to improve the operational picture, they seem to want citizens to use SM during an emergency. It would be useful for the ES if the citizens are posting information the ES needs but does not have – especially in the beginning of a crises, when the ES is still on its way to the incident. Information validation How can emergency services be persuaded that information provided via social media from citizens is credible and trustworthy?
The interviewees were very worried about credibility. They said that the ES are very interested in pictures from the scene, but that “pictures can be faked very easily” – a major issue. But they also said, that somebody living in the area of the incident, will be able to distinguish between fake and real very fast. E.g. during the Munich shooting in 2016 fake pictures from a different country were spread through different channels. Due to the fact, that the policemen knew the site of the shooting very well, they were able to identify the fake pictures very fast.
It would be helpful to have people working on SM from all over Graz. The interviewees said, that it should even be possible to get these people, if the members of the ES itself would become a kind of ‘VOST’- team. For they are living all over Graz, there should be always someone, “who knows the scene of the incident well enough to say true/fake”. Another approach to this issue was seen in the use of the combination of other information sources: “One picture alone would not be enough”. What procedures and tools are used to validate such information efficiently and effectively? How can reliability and accuracy be supported, technically?
-Example of Information Validation Emergency service staff skills and resources
One of the employees responsible for managing the SM communication has studied journalism and public relations. This was important to be fast and efficient in terms of coordination and communication. What skills are required to help emergency service staff to find, interpret and make use of social media data?
-What guidance or training would be most useful for them?
197/211 D2.3: Impact of social media, Version V1, PU Moderating citizen communities Key Conclusions
After the Graz incident the awareness for the use of SM has risen. Now there are plans to create and establish a SM-plan and –strategy. There is also the possibility to try to win the staff of the city over to the VOST. This could be very helpful as they work and live in different places all over the city. Further sources relevant for this case study
http://www.graz.at/cms/beitrag/10251761/1618648 https://en.wikipedia.org/wiki/2015_Graz_van_attack http://www.thelocal.at/20150623/police-reconstruct-scene-of-graz-tragedy
Pictures and videos: http://www.bbc.com/news/world-europe-33211479 http://www.dailymail.co.uk/news/article-3132535/Man-runs-car-crowd-killing-2-injuring-Austria.html
198/211 D2.3: Impact of social media, Version V1, PU Appendix 9: Significance Tables Table 1: Chi-square tests statistics
Variable Variable 2 χ2 df N p Cramer’s 1 V Country Use of social media in emergencies 239.57 9 6723 < .110 0.001 Country Weather conditions or warnings 70.07 3 1983 < .188 0.001 Country Road and traffic conditions 97.91 3 1982 < .222 0.001 Country Your feelings or emotions about what 21.43 3 1982 < .104 was happening 0.001 Country Your location 10.71 3 1981 0.013 .074 Country What actions you were taking to stay 20.72 3 1981 < .102 safe 0.001 Country An eyewitness description of 32.13 3 1981 < .127 something you experienced 0.001 Country An eyewitness photo 183.84 3 1983 < .304 0.001 Country A video 76.01 3 1983 < .196 0.001 Country Other 22.41 3 1983 < .106 0.001 Country Reassurance that you are safe 7.01 1982 0.072 .059 Country Advice about what actions others 4.23 3 1983 0.237 .046 should take to stay safe Country Having downloaded an app 406.11 6 7071 < .170 0.001 Country Weather app 29.38 3 1065 < .159 0.001 Country Warning app 23.63 3 1067 < .142 0.001 Country Emergency call up 30.29 3 1066 < .161 0.001 Country Other app 8.33 3 1065 < .085 0.001 Country First Aid app 6.84 3 1065 .077 .077
199/211 D2.3: Impact of social media, Version V1, PU Table 2: One-way ANOVA test statistics
2 Independent Dependent F dfbetween dfwithin p η variable variable Country Emergency 147.72 3 7230 < 0.001 .058 services should regularly monitor their social media Country I would expect to 131.53 3 7230 < 0.001 .052 get a response from them within an hour Country Emergency 46.52 3 7230 < 0.001 .019 services are too busy to monitor social media during an emergency
200/211 D2.3: Impact of social media, Version V1, PU Table 3: Kruskal-Wallis test statistics
Independent Dependent χ2 df N p variable variable Country Information on 147.56 3 7234 < 0.001 social media is not reliable Country There are many 156.62 3 7234 < 0.001 false rumours on social media Country I am concerned 126.12 3 7234 < 0.001 about data privacy Country It is better to call 131.06 3 7234 < 0.001 999 than to post messages on social media Country I am not confident 59.17 3 7234 < 0.001 using social media Country Social media might 111.15 3 7234 < 0.001 not work properly in an emergency Country receive emergency 514.44 3 7234 < 0.001 warnings Country find out 381.54 3 7234 < 0.001 information about the emergency Country receive tips about 367.37 3 7234 < 0.001 how to stay safe Country connect with other 447.00 3 7234 < 0.001 citizens to help others affected by the emergency Country contact an 417.40 3 7234 < 0.001 emergency service instead of making a 999 call’ Country share information 345.83 3 7234 < 0.001 about the emergency with an emergency service
201/211 D2.3: Impact of social media, Version V1, PU Appendix 10: Full tables - Revised explanation tables
Table 1: Alternative Primary Explanation Mechanisms M1A (2)
Alternative M1A: Citizens and ES less strongly connected through social media Explanation
Mechanism Citizens and ES increased awareness of each other’s activities leads to fragmentation of behaviours, sometimes in oppositional ways to those intended
Type Direct Rival
Level Individual
Identifiers London Riots case study shows social media supported the spread of the riots themselves and helped coordinate activities of rioters.
Flooding case studies show social media plays complementary or subordinate role to conventional media. Flooding case studies show different levels of co-operation between citizens and ES, with some ES adopting ‘top-down’ approach; some citizens not sharing with ES. Flooding cases show although many citizens shared information, many ES did not use it because of access issues, information overload or validity issues.
The third wave of case studies and the Citizens survey showed the main disadvantage of relying heavily on Twitter and Facebook as sources of social media information and channels of communication is that they ignore other social media platforms growing in popularity with particular demographics (like Snapchat and Instagram) and increasingly have less access to certain groups like young people (who tend to use closed source channels like WhatsApp). This leads to fragmentation of channels of communication between different citizen groups and ES.
However, the third wave of case studies show that the impact of social media in emergencies is enhanced when social media is incorporated within broader systems and processes aimed at supporting transparency and open government. In Eindhoven and Rotterdam, social media has contributed to establishing a culture of ‘information open- ness’ which in turn has laid the foundations for establishing the municipalities, and their public safety and emergency services, as credible and trusted sources of information for citizens and other stakeholders. This has consequently increased the motivation of citizens and other stakeholders to play an active role as information providers
and validators. Moreover, the widespread engagement of citizens has reduced the need to cultivate ‘citizen communities’ as part of the emergency management system.
Degree of Influence
202/211 D2.3: Impact of social media, Version V1, PU
Certainty Moderate. This patterns broadly fits the alternative explanation
Robustness Low. Low triangulation with other data sources. Conflicting explanations between wave 2 and wave 3 case studies.
Range Low. The behaviours cited are only two of a large range of contributory factors
Prevalence Low. The explanation relates to sub-groups of the target population
Grounding Low. There is very little evidence in the literature to support this explanation
Implication The evidence that the use of social media weakens the connection between citizens and emergency services is weak. However, the evidence does suggest that neither citizens nor emergency services are homogenous groups. Citizens take up different roles during emergencies, including informant, helper, amplifier and citizen journalists - and they also use different technologies. Particular types of social media are more likely to be used by certain demographic and socio-economic profiles for particular purposes. Emergency services differ in their social media skills and knowledge as well as their organisational context. Different types of organisation adopt different bridging and engagement strategies with citizens. This implies that the likely impact of social media will vary depending on the cultural and organisational context in which it is used.
The evidence shows that ES and EmerGent both need to take account of changes in social media technologies, which reflect greater differentiation in the use of different services by different groups for different purposes.
203/211 D2.3: Impact of social media, Version V1, PU Table 2: Alternative Primary Explanation Mechanisms M2A (2)
Alternative M2A: Parallel lines Explanation
Mechanism Social media runs as a parallel activity within ES systems, and is utilised by individuals or small groups as a ‘bonus feature’ without being integrated, thereby restricting the capacity of ES to respond more effectively to incidents and leading to citizens not fully understanding the benefits of using social media in emergencies
Type Direct Rival
Level Institutional-interpersonal
Identifiers The flooding case studies show that many emergency services did not have the skills and resources to use information from social media effectively during the floods.
Flooding case studies show lack of organisational structures and procedures in place to support the use of social media.
The Wave 3 case studies show a more broader, and more sophisticated, adoption of social media tools than was indicated in previous case study waves, with several examples of ES incorporating data management tools – like Coosto, Tweekdeck and Hoostsuite - within their existing systems. These tools are used in different ways, according to the priorities, target groups and strategies of the organisation, and are contextualised for purpose. In most case studies, the use of social media is integrated into the emergency management cycle and that use of social media is an integral / routine part of their work and is supported by senior management, or at a higher political level.
Degree of Influence
Certainty Moderate. This patterns broadly fits the alternative explanation
Robustness Low. Low triangulation with other data sources
Range Low. The pattern shown is only one of a large range of contributory factors
Prevalence High. Case studies show this pattern was repeated across different ES
Grounding Low. There is very little evidence in the literature to support this explanation
Implication The introduction of social media into ES organisational systems and practices can be disruptive. To strengthen connections between social media tools and ES systems requires relevant skills and new organisational systems and processes. There is some evidence that some emergency services will be resistant to these disruptive changes, at some levels,
204/211 D2.3: Impact of social media, Version V1, PU leaving social media as an ‘outsider’ or ‘exotic’ feature that is utilised only by social media ‘champion’s and innovators.
The wave 3 case studies showed clearly how the organisational culture of a service will shape how social media tools are embedded in the organisations’ emergency management processes / procedures. The case studies suggested a typology of ‘whole system enthusiasts’, ‘bottom up enthusiasts’, ‘top down professionals’ and ‘worried pragmatists’. EmerGent needs to ensure that the EIS, App and Guidelines reflect this typology.
205/211 D2.3: Impact of social media, Version V1, PU Table 3: Alternative Primary Explanation Mechanisms M3A (2)
Alternative M3A: Social media analysis tools contribute to system dysfunctionality Explanation
Mechanism Added flow of information overloads ES IT systems and hampers functionality so that situational awareness is decreased
Type Direct rival
Level Institutional
Identifiers London riots case study shows turning information into intelligence is more challenging during extreme events resulting in millions of tweets and other social media data in a short period of time
Flooding case studies showed instances where social media information overload contributed to reduced functionality and integration of ES IT systems.
The wave 3 case studies showed that information overload per se is not the major impediment to improved situational awareness. Two major problems encountered by ES in handling social media information were, first, identifying the geographical location of messages and, second, false positives - despite the use of complex search strings and other functions, most tools still provide a lot of information that is not relevant to a particular emergency or service; this means that during a major incident those using such a tool may struggle to identify relevant messages or resort to searching by hashtags (#) only which can result in missing other important information not shared in this way.
Another consistent theme across all of the case studies in wave 3 was the role traditional media – and journalists – play in emergency situations and the importance of integrating this human element into systems aimed at improving situational awareness.
Degree of Influence
Certainty Moderate. This patterns broadly fits the alternative explanation
Robustness Moderate. Triangulation of data sources show recurring pattern
Range Low. The pattern shown is only one of a large range of contributory factors
Prevalence High. Case studies show this pattern was repeated across different ES
Grounding Low. There is very little evidence in the literature to support this explanation
206/211 D2.3: Impact of social media, Version V1, PU
Implication There is some evidence to cast doubt on EmerGent’s third expected (desired) outcome – that the Social App, EmerGent IT and ES IT are better integrated because, by integrating social media into the ES workflow, social media data are used more often during emergencies and more information is posted by ES; functionality for assessing a situation is better integrated and situational awareness improves. What seems to be a repeated pattern is that integrating information from social media apps and other sources within ES control systems can disrupt ‘normal’ information flow and lead to system overload. This could lead to the opposite result to that intended – a situation in which situational awareness in ES control rooms is undermined.
However, wave 3 of the case studies highlighted two key issues that the EmerGent interface would need to address in order to improve situational awareness – identifying the geographical source of information, and reducing the amount of irrelevant information gathered. The EmerGent system needs to be able to handle search strings to strike a balance between general categories (e.g. capturing information about ‘fires’ ‘floods’ etc.) but also to reflect local contextual conditions, particularly the geography of the area in which the tools are being used.
The case study analysis shows that integration of social apps, IT systems and ES IT systems needs to also involve integration of social media with traditional media – and in particular creating the conditions whereby social media experts work with journalists.
207/211 D2.3: Impact of social media, Version V1, PU Table 4: Influencing Factor – Cultural Contextualisation (2)
Influencing Factor Cultural contextualisation
Mechanism Familiarity with and routine embedding of social media in individual daily life will increase the probability that citizens will use social media to engage with other citizens and ES in emergencies and will shape their use behaviours
Type Implementation rival
Level Interpersonal
Identifiers Flooding case studies show that social media are used for different purposes by different people, and have cultural specificities (‘cultural protocols’). Moreover, the trajectory of their development is shaped by use (‘value embedded action’). It also matters whether social media are open or closed, and how people view them depends very much on who they are choosing to follow
Flooding case studies show that a key variable determining use of social media in emergencies is already using it in normal daily life.
The Wave 3 case studies – particularly in Rotterdam and Eindhoven – showed that when social media is embedded as routine practice as part of an ‘open government’ approach it increases the engagement of citizens as co-producers of information in emergencies.
Degree of Influence
Certainty Moderate. The case study evidence, supported by the literature review and a 2014 survey on citizens use of social media, shows that regular users of social media are more likely to use it in emergencies.
Robustness Moderate. The mechanism was identified as contributing factor in literature reviews, the Citizens Survey and the case studies.
Range Moderate. The explanation only accounts for one possible behavioural factor in citizens use of social media in emergencies
Prevalence Moderate. This factor was consistently identified across a wide range of locations in the three waves of case studies.
Grounding Moderate. There is some evidence from ‘value embedded action theory’ (Cullen and Cohen, 2007) that technologies have cultural specificities and that the trajectory of their development is shaped by use (‘value embedded action’).
Implication The evidence suggests that EmerGent is most likely to achieve its intended outcome of developing a stronger connection between citizens and emergency services by improving
208/211 D2.3: Impact of social media, Version V1, PU citizens’ awareness of emergency services information needs (both in terms of content and quality) among those already using social media. It also implies that that the strategies ES devise to maximise the potential of social media need to reflect a typology of different cultural protocols. Conversely, ES cannot assume that all citizens uniformly will respond in the same way to the same messages. Another implication is that if the EmerGent guidelines are going to effect change, they have to cater for a range of experiences and needs. Finally, in order to help emergency services use the tools provided by EmerGent they need to be provided with help and guidance on how to do this.
209/211 D2.3: Impact of social media, Version V1, PU Table 5: Influencing Factor – Organisational Culture (2)
Influencing Factor Organisational culture
Mechanism The impact of SM information on ES management of emergencies depends on key organisational culture factors like the efficiency of ES info-management systems; skills base and experience of using social media
Type Implementation rival
Level Institutional
Identifiers Flooding case studies show that efficient reach for ES with social media comes with the trade-off of receiving greater citizen responses which can place an additional burden on emergency services – the way they respond then affects the mood of citizens.
The London Riots and the Flooding case studies show that ES often lack the practice and skills of identifying and interpreting relevant information.
Third wave case studies show a key training need now in emergency services and similar agencies is to keep up with trends in social media.
The ES Surveys showed that 86% of staff thought that acquiring skills to use social media analysis tools was very important.
The case studies of the field trials showed that, what functionalities are needed and how the IES tool would be used depends on the target group.
The ES Surveys showed that 84% of staff thought that organisational culture was important in ensuring social media was used effectively.
Both Dortmund and Hamburg trials suggest that introducing the system into an emergency services organisation is likely to require some training.
Both Dortmund and Hamburg trials suggest that there is likely to be a need for local adaptation of the EmerGent suite of tools so that usage is in line with organisational processes and practices.
The case studies of the field trials show that keywords need to be adjusted (continuously) to a particular context
Degree of Influence
Certainty High. Three waves of case study evidence shows this pattern is found across a range of emergency services.
Robustness Moderate. The mechanism was identified as contributing factor in literature reviews, an ES Survey and the case studies.
210/211 D2.3: Impact of social media, Version V1, PU
Range High. The data suggest consistently that organisational culture is a key factor in ES use of social media in emergencies
Prevalence High. Successive waves of case study and survey data, plus demos and field trials provides high degree of triangulation that this factor occurs often.
Grounding Moderate. There is an established body of evidence form organisational theory that supports the concept of organisational resistance.
Implication The evidence suggests that the use of social media is not yet common practice for all emergency services and they often lack the practice and skills of identifying and interpreting relevant information. This implies that the scale of EmerGent impact in relation to emergency services broadcasting use of social media is likely to vary depending on existing organisational practices. One of the key challenges emergency services have at present is accessing and making sense for response purposes of the increased volume of social media information that is generated during emergencies. In none of our case studies did emergency services have a way to systematically access social media data. It would therefore seem that here lies a great potential impact for EmerGent.
211/211